RESULTS AND DISCUSSION

To develop a positive selection assay for protein degraders, we made a bicistronic lentivirus encoding (i) a POI fused to a modified version of deoxycytidine kinase (hereafter called DCK*) that converts the non-natural nucleoside 2-bromovinyldeoxyuridine (BVdU) into a poison (4) and (ii) green fluorescent protein (GFP). We reasoned that GFP could be used to mark reporter-positive cells, to FACS (fluorescence-activated cell sorting) sort for cells with the desired reporter mRNA levels, and to count cells in multiwell plate assays. In our initial proof-of-concept experiments, we used this virus to create 293FT cells expressing the IMiD target IKZF1 (1, 2, 5) fused to DCK* and compared them to cells expressing unfused DCK* or unfused IKZF1 (Fig. 1, A and B). As expected, the IMiD pomalidomide (POM) down-regulated DCK*-IKFZ1 and IKZF1 but not DCK* (Fig. 1B). We also confirmed that 293FT cells expressing DCK*-IKZF1 or unfused DCK* were more sensitive to BVdU than 293FT cells expressing IKZF1 alone or infected with an empty vector (EV) (Fig. 1C). The increased BVdU sensitivity of the DCK* cells relative to the DCK*-IKZF1 cells is likely explained by the higher protein levels of DCK* compared to DCK*-IKZF1 (Fig. 1B). Similar results were observed with cells expressing DCK*-K-Ras (G12V), DCK*-Cyclin D1, DCK*-FOXP3, and DCK*-MYC, indicating that DCK* remains active when fused to a variety of proteins (fig. S1). POM increased the BVdU median effective concentration (EC50) of cells expressing DCK*-IKZF1 but not of cells expressing DCK* (Fig. 1D).

(A) Vector schematic. DCK*, variant deoxycytidine kinase with Ser74Glu, Arg104Met, and Asp133Ala substitutions; V5, V5 epitope tag; GGS, Gly-Gly-Ser spacer; IRES, internal ribosomal entry site. (B) Immunoblot analysis of 293FT cells infected with the lentiviral vectors depicted in (A) and then treated with 1 or 10 M POM, as indicated by the triangles, for 24 hours. (C and D) Relative survival of 293FT cells infected with the lentiviral vectors depicted in (A) and then treated with the indicated concentrations of BVdU for 4 days. In (D), cells were also treated with 1 M (POM) starting 24 hours before BVdU was added. n = 3 biological replicates. (E and F) Number of GFP-positive 293FT cells infected to produce DCK* (E) or DCK*-IKZF1 (F) using the vectors in depicted in (A) and then treated with indicated concentrations of POM and BVdU in 384-well plate format. POM was added 24 hours before treatment with BVdU for 4 days. n = 4 biological replicates. (G) Immunoblot analyses of cells treated as in (E) and (F). (H) Fluorescence data of 384-well plate containing 293FT cells expressing DCK*-IKZF1 treated with DMSO (columns 1 to 11 and 24) or 1 M POM (columns 12 to 23), followed 24 hours later by the addition of 100 M BVdU for 4 days (columns 1 to 24).

Next, we seeded either the DCK*-IKZF1 cells or DCK* cells in 384-well plates and treated the wells with increasing amounts of POM or with dimethyl sulfoxide (DMSO). We added BVdU 24 hours later and measured cell viability 4 days thereafter by measuring the number of GFP-positive objects per well. POM again promoted the survival of the DCK*-IKZF1 cells, but not the DCK* cells, over a range of POM and BVdU concentrations (Fig. 1, E to G). In anticipation of using this assay for a high-throughput screen, we next seeded the DCK*-IKZF1 cells in 384-well plates and treated half the wells with POM and half the wells with DMSO, followed 24 hours later by BVdU (Fig. 1H). Measuring GFP-positive objects 4 days later produced a favorable Z value (0.7) for this assay.

Encouraged by these findings, we did a pilot screen with 293FT cells expressing DCK*-IKZF1 or unfused DCK* grown in 384-well plates and a library of ~2000 bioactive compounds, which included lenalidomide (LEN) and POM (Fig. 2, A to C). Each well received a different compound at a concentration of approximately 10 M by pin transfer, followed the next day by BVdU. BVdU was added at 100 M to the DCK*-IKZF1 cells and at 10 M to the DCK* cells to achieve comparable cell killing despite the higher levels of DCK* relative to DCK*-IKZF1 (fig. S2). Four days thereafter, the GFP fluorescence for each well was measured and converted to a z score based on the GFP fluorescence values for the other wells on its plate. LEN and POM scored positively (z > 2) in the DCK*-IKZF1 screen but not the DCK* screen (Fig. 2, B to E). Some compounds promoted the survival of both DCK* cells and the DCK*-IKZF1 cells, including compounds that interfere with BVdU uptake (e.g., dipyridamole) (6, 7) or incorporation into DNA (e.g., thymidine) (compare Fig. 2, B and C). Such assay positives could be largely eliminated by subtracting the DCK* z score for each chemical from its DCK*-IKZF1 z score (Fig. 2F). For comparative purposes, we also did a screen with the same 2000 bioactive compound collection using 293FT cells expressing a bicistronic mRNA encoding (i) an IKZF1Firefly luciferase (Fluc) fusion and (ii) Renilla luciferase (Rluc), using a decrease in the Fluc/Rluc ratio to identify IKZF1 degraders (Fig. 2, G to I, and fig. S3) (2). As expected for such a down assay, this screen underperformed the DCK*-IKZF1 up screen with respect to both signal to noise and the number of false positives, which included compounds that inhibit Cap-dependent translation (e.g., VX-11e or BIX02565) (810). Compounds that nonselectively inhibit transcription, translation, or protein folding would predictably be especially problematic for Fluc fusions with shorter half-lives than the Rluc internal control. Notably, the transcriptional inhibitor actinomycin D and the translational inhibitor cycloheximide did not promote the survival of the DCK*-IKZF1 cells at any concentration tested (fig. S4).

(A) Scheme for positive selection protein degradation assay. (B and C) Representative fluorescence data of 384-well plates containing 293FT cells expressing DCK* (B) or DCK*-IKZF1 (C) treated with compounds in the Selleck BioActive Library (one compound per well), followed 24 hours later by the addition of BVdU at the EC85 (10 and 100 M, respectively) for 4 days. BVdU was omitted in column 1. Columns 23 and 24 contained 10 M POM and 12.5 M dipyridamole (DiP), respectively. Library wells containing POM and DiP are indicated by the red and white arrows, respectively. (D and E) Z-distribution of GFP fluorescence of DCK* cells (D) and DCK*-IKZF1 cells (E) screened with the full Selleck BioActive Library. LEN and POM are indicted by the blue circle and red triangle, respectively. n = 2 biological replicates. (F) Corrected z scores obtained by subtracting z scores in (D) from z scores in (E). (G) Scheme for negative selection screening using the dual-luciferase reporter assay. (H and I) Z scores of Fluc/Rluc ratio of 293FT IKZF1-Fluc-IRES-Rluc cells after screening with the Selleck BioActive Library for 8 hours (H) or 4 days (I). n = 2 biological replicates.

As one way to minimize false positives, we seeded 384-well plates with a 1:1 mixture of 293FT cells expressing either (i) DCK*-IKZF1 and GFP or (ii) DCK* and TdTomato (Fig. 3A). Both POM and dipyridamole increased the number of GFP-positive cells, but dipyridamole was readily identified as a false positive by examining the TdTomato fluorescence channel (Fig. 3B). We then repeated these experiments in 384-well plate format, exposing the cells to 10 different concentrations of a small library of approximately 100 analogs of POM that we had synthesized, which included the known IKZF1 degraders LEN, POM, and avadomide (MI-2-65) (11) and several uncharacterized IMiD-like molecules from the literature (12) (Fig. 3C and tables S1 and S2). This library was generated to test whether our assay could correctly identify the known IKZF1 degraders and identify additional IKZF1 degraders made by alternative diversification of the aryl moiety of POM. LEN, POM, and avadomide all scored in our assay (Fig. 3C). In addition, several previously uncharacterized compounds, including MI-2-61 and MI-2-197, appeared to be at least as potent as POM in this screen and in confirmatory immunoblot assays (Fig. 3, C to F, and fig. S5). Our screen also correctly classified compounds that did not down-regulate IKZF1 in immunoblot assays, including some (e.g., MI-2-192 and MI-2-118) that still bound to cereblon in biochemical assays (fig. S5).

(A) Scheme for in-well GFP/TdTomato competition assay. 293FT cells were infected to produce DCK*-IKZF1 and GFP or DCK* and TdTomato using bicistronic vectors analogous to those depicted in Fig. 1A. (B) Top: Heatmap of the fold change (relative to treatment with DMSO) of GFP fluorescence of a 1:1 mixture of GFP-positive DCK*-IKZF1 and TdTomato-positive DCK* cells treated with 3.125, 6.25, 12.5, or 25 M POM or dipyridamole or with vehicle (DMSO) and followed 1 day later by the addition of 100 M BVdU for 4 days. Bottom: Heatmap of the fold change (relative to treatment with DMSO) of the ratio of GFP fluorescence to TdTomato fluorescence of the cells treated in (A). n = 2 biological replicates. (C) Heatmap of the fold change (relative to treatment with DMSO) of the ratio of GFP to TdTomato fluorescence of a 1:1 mixture of GFP-positive DCK*-IKZF1 and TdTomato-positive DCK* cells treated with 1.3 nM, 3.8 nM, 11.4 nM, 34 nM, 102 nM, 310 nM, 920 nM, 2.78 M, 8.33 M, and 25 M of the indicated IMiDs, as indicated by the triangles, or with vehicle (DMSO), and followed 1 day later by the addition of 100 M BVdU for 4 days. n = 2 biological replicates. (D) Immunoblot analysis of 293FT cells lentivirally transduced to express IKZF1-V5 and treated with the indicated IMiD derivatives for 24 hours using the same concentration range as in (C). (E) Structures of POM and IMiD MI-2-61. (F) Quantification of immunoblot data in (D); n = 2 biological replicates.

To begin looking for non-IMiD IKZF1 degraders, we screened ~546 metabolic inhibitors and anticancer drugs at 10 different concentrations using the DCK*-IKZF1 293FT cells in 384-well plate format (tables S3 and S4) (13). In parallel, we counterscreened against unfused DCK* cells. Spautin-1 (14), like POM, promoted the survival of the DCK*-IKZF1 cells, but not the DCK* cells, in a dose-dependent manner (Fig. 4, A to D). We confirmed that Spautin-1 down-regulated DCK*-IKZF1 and V5-tagged exogenous IKZF1 but not DCK* (Fig. 4E). IKZF1-V5 was among the 100 most down-regulated proteins after 24 hours of Spautin-1 treatment, as determined by quantitative mass spectrometry proteomics (fig. S6 and table S5). Until the direct target of Spautin-1 linked to IKZF1 turnover is known, it is impossible to know how many of these changes in protein abundance are direct versus indirect and on-target versus off-target. Notably, Spautin-1, unlike POM, down-regulated IKZF1 in cells lacking cereblon (Fig. 4F).

(A) Chemical structure of Spautin-1. (B) GFP fluorescence of DCK*-IKZF1 and DCK* 293FT cells treated with ranolazine, Spautin-1, and resveratrol at concentrations of 25 M, 8.33 M, 2.78 M, 920 nM, 310 nM, 102 nM, 34 nM, 11.4 nM, 3.8 nM, and 1.3 nM, as indicated by the triangle, followed 24 hours later by the addition of BVdU at the EC85. Shown for comparison are cells treated with POM (10 M) or dipyridamole (DiP) (12.5 M) before adding BVdU. n = 2 biological replicates. (C and D) Quantification of GFP fluorescence from (B) for Spautin-1 (C) and for an analogous titration with POM (D). (E) Immunoblot analysis of 293FT cells infected with lentiviruses as in Fig. 1A and treated with the indicated concentrations of Spautin-1 for 24 hours. (F) Immunoblot analysis of isogenic 293FT CRBN +/+ and CRBN / cells transduced to express IKZF1-V5 and treated with the indicated concentrations of Spautin-1 or POM (1 M) for 24 hours. (G) Immunoblot analysis of 293FT cells stably expressing IKZF1-V5 and simultaneously treated with MLN7243 (1 M), MLN4924 (1 M), MG132 (1 M), Spautin-1 (10 M), or POM (1 M) for 24 hours as indicated. (H and I) Immunoblot (H) and RT-qPCR (I) analysis of KMS11 multiple myeloma cells treated with indicated concentrations of Spautin-1 or POM (1 M) for 24 hours. n = 3 biological replicates.

Spautin-1 reportedly suppresses autophagy by inhibiting the USP10 and USP13 deubiquitinases (14). IKZF1 protein levels were not decreased after small interfering RNAmediated down-regulation of USP10, alone or in combination with USP13 (fig. S7A), and Spautin-1s ability to down-regulate IKZF1 was not altered when one or both of these proteins were suppressed (fig. S7, B and C). Moreover, Spautin-1 down-regulated IKZF1 in 293FT cells in which autophagy was disabled by CRISPR-Cas9mediated disruption of ATG7, Beclin1, or FIP200 (fig. S8).

In contrast, down-regulation of IKZF1 by Spautin-1 was blocked by compounds that inhibit either the E1 ubiquitin activating enzyme or the proteasome (Fig. 4G). Down-regulation of IKZF1 by Spautin-1 was not, however, blocked by an inhibitor of neddylation, which is required for cullin-dependent ubiquitin ligases [e.g., the cereblon-containing ubiquitin E3 ligase that is coopted by the IMiDs (1, 2, 5)] (Fig. 4G). Down-regulation of exogenous IKZF1 by Spautin-1 requires the IKZF1 N-terminal region containing IKZF1s first zinc finger domain (ZF1) but not the IKZF1 zinc finger domain (ZF2) targeted by the IMiDs (fig. S9, A and B) (15, 16). The down-regulation of the N terminus of IKZF1 was similarly blocked by compounds that inhibit either the E1 ubiquitin activating enzyme or the proteasome but not by inhibitors of neddylation (fig. S9C). Preliminary structure-activity relationship studies identified both active and inactive Spautin-1 derivatives (fig. S10), suggesting that down-regulation of IKZF1 by Spautin-1 reflects a specific protein-binding event and that Spautin-1s potency and specificity can be optimized further.

The experiments described above implied that Spautin-1 posttranscriptionally regulates IKZF1. Nonetheless, Spautin-1 also suppressed exogenous IKZF1 mRNA levels in 293FT cells (fig. S11). However, Spautin-1 suppressed endogenous IKZF1 protein levels in KMS11 and L363 myeloma cells at concentrations that minimally suppressed IKZF1 mRNA levels (Fig. 4, H and I, and fig. S12, A to C). Spautin-1 did not down-regulate IKZF1 in all myeloma cells tested (fig. S12, D and E). The biochemical basis for this variability is not clear.

Notably, down-regulation of IKZF1 by Spautin-1 occurs much more slowly than with IMiDs, suggesting that its effect on IKZF1 is indirect (fig. S13). Nonetheless, its ability to score in a positive selection assay, as well as its inability to down-regulate IKZF1 in some myeloma lines, suggests that it is not broadly toxic at concentrations that down-regulate IKZF1. We are currently seeking the direct Spautin-1 target linked to IKZF1 turnover using genetic and biochemical tools.

One advantage of positive selection assays is their enablement of pooled screens. Our positive selection assay, however, uses a suicide gene. Some suicide genes cause bystander killing that could confound their use in pooled screens. In pilot studies, however, we confirmed that DCK*-IKZF1 cells rapidly outgrew DCK* cells in cocultures treated with IMiDs and BVdU (fig. S14A) and that the DCK* single guide RNA (sgRNA) was rapidly and specifically enriched relative to the control sgRNA in Cas9-positive 293FT cells expressing either DCK*-IKZF1 or DCK*-FOXP3 and then treated with BVdU (fig. S14, B and C). Therefore, bystander killing is negligible in this system.

To begin to address the general utility of our methodology, as well as its ability to function in a pooled format, we next did experiments with ASCL1 in place of IKZF1. ASCL1 is an undruggable lineage-specific transcription factor that is required for survival in many small cell lung cancers (SCLCs) and neuroblastomas (1719). We made Jurkat T cells that express Cas9 and either (i) DCK*, (ii) the neural/neuroendocrine lineagespecific transcription factor ASCL1, (iii) DCK*-ASCL1, or (iv) ASCL1-DCK* (Fig. 5A and fig. S15A). Jurkat cells were chosen because they are easily grown and expanded in suspension cultures. ASCL1-DCK* was chosen for further study because we could not generate cells producing high levels of DCK*-ASCL1 (fig. S15A). We first confirmed that ASCL1-DCK* expression sensitized Jurkat cells to BVdU and that this was partially reversed after down-regulating the fusion with ASCL1 sgRNAs (Fig. 5, A and B, and fig. S16, A and B). Cas9 expression was also slightly attenuated in the ASCL1-DCK* cells over time for unclear reasons (Fig. 5A). Nonetheless, these cells efficiently edited a GFP-based reporter of Cas9 activity within 10 days of receiving a GFP sgRNA (fig. S15, B and C). Next, we infected the ASCL1-DCK* and DCK* cells with a lentiviral sgRNA library targeting 788 genes (seven sgRNAs per gene) that encode druggable proteins (table S6). Ten days later (to allow time for gene editing), the cells were split and grown in the presence of 200 or 500 M BVdU for an additional 2 weeks (fig. S16C). We then determined sgRNA abundance by next-generation sequencing of genomic DNA and analyzed relative enrichment of sgRNAs compared to the time point before BVdU treatment (fig. S16D). We identified multiple sgRNAs against CDK2 that were markedly enriched at both BVdU concentrations in the ASCL1-DCK* cells but not the DCK* cells (Fig. 5C, fig. S16, D and E, and table S7).

(A) Immunoblot analysis of Jurkat cells first infected to express Cas9 and then superinfected to express exogenous ASCL1, DCK*, or the ASCL1-DCK* fusion. NCI-H69 cells are included as a benchmark for ASCL1 endogenous expression. (B) Growth inhibition (%), based on viable cell numbers relative to untreated controls, of the indicated cell lines from (A) treated with BVdU for 6 days. n = 2 biological replicates. (C) Hypergeometric analysis of BVdU positive selection CRISPR-Cas9 screen on day 25 relative to day 10 (early time point before BVdU treatment) of ASCL1-DCK* Cas9 Jurkat cells treated with 500 M BVdU. n = 2 biological replicates. (D) Quantification of fold change in mCherry:BFP ratio after 18 days of 500 M BVdU or DMSO (0) treatment of ASCL1-DCK* Cas9 Jurkat cells expressing the indicated sgRNAs and mCherry or a nontargeting control sgRNA and blue fluorescent protein (BFP) (initially mixed 1:3). n = 3 biological replicates. (E) Immunoblot and (F) RT-qPCR analysis of ASCL1-DCK* Cas9 Jurkat cells superinfected to express the indicated sgRNAs. n = 4 biological replicates. (G) Immunoblot analysis of Jurkat cells first infected with a lentivirus to stably express exogenous ASCL1, then infected with Dox-inducible (DOX-On) sgRNA-resistant CDK2 wild-type (WT) or CDK2 kinase-dead (KD) mutant, and lastly superinfected with a CDK2 or nontargeting sgRNA. Following superinfection with the sgRNA lentiviruses, cells were grown in DOX to maintain exogenous CDK2 expression. n = 4 biological replicates. Exo, exogenous CDK2; Endo, endogenous CDK2. Error bars represent SD. ns, nonsignificant; *P < 0.05; ***P < 0.001; ****P < 0.0001.

In validation studies, ASCL1-DCK* Jurkat cells expressing CDK2 sgRNAs outcompeted ASCL1-DCK* cells expressing control sgRNAs in the presence of BVdU but not in the presence of DMSO (Fig. 5D and fig. S16F). CDK2 sgRNAs also posttranscriptionally down-regulated ASCL1-DCK* protein levels in the Jurkat cells (Fig. 5, E and F) and endogenous, unfused, ASCL1 in human SCLC lines (NCI-H1876 and NCI-H2081) (Fig. 6, A and B, and fig. S18, A and B). Down-regulation of exogenous ASCL1 in Jurkat cells treated with a CDK2 sgRNA was rescued by an sgRNA-resistant CDK2 complementary DNA (cDNA) encoding wild-type CDK2 but not kinase-dead CDK2 (Fig. 5G). The kinase-dead CDK2 was, however, produced at slightly lower levels, presumably because it is less stable or because of its known dominant-negative effects due to cyclin sequestration (2022).

(A) Immunoblot and (B) RT-qPCR analysis of the NCI-H1876 SCLC cell line that endogenously expresses ASCL1 infected to express the indicated sgRNAs. n = 3 biological replicates. (C and E) Immunoblot and (D and F) RT-qPCR analysis of NCI-H1876 human SCLC cells (C and D) and 97-2 mouse SCLC cells (E and F) after treatment with the CDK2 PROTAC degraders (TMX-2138 and TMX-2172) or the indicated negative controls, all used at 500 nM for either 36 hours (C and D) or 8 hours (E and F). Neg Deg, negative control degrader ZXH-7035. n = 3 biological replicates. (G) Immunoblot analysis and (H) quantification of ASCL1 protein levels in 97-2 cells first treated with the CDK2 PROTAC degrader or negative control (500 nM) for 4 hours and then treated with cycloheximide (CHX) (150 g/ml) for the indicated times. S.E., short exposure; L.E., long exposure. n = 4 biological replicates. In all experiments, error bars represent SD except in (H), where error bars represent SEM. *P < 0.05; ***P < 0.001; ****P < 0.0001.

CDK2 has been well recognized as a potential anticancer target. The development of selective small-molecule CDK2 inhibitors, however, has been hampered by their off-target effects on other CDK family members, especially the broadly essential kinase CDK1. We verified that well-established CDK2 inhibitor dinaciclib (23) down-regulated both ASCL1 protein and mRNA levels (fig. S17, A to D), potentially due to its polypharmacological activity on both CDK2 and other CDKs such as CDK9 (23, 24). We obtained, however, two small-molecule CDK2 degraders (TMX-2138 and TMX-2172) that more selectively target CDK2 through recruitment of cereblon (25). Both of these compounds down-regulated ASCL1 protein levels in both human (NCI-H1876 and NCI-H1092) and mouse (97-2 and 188) SCLC lines (Fig. 6, C to F, and fig. S18, C to F). For unclear reasons, ASCL1 was down-regulated more rapidly in the mouse lines than in the human lines. We focused on TMX-2172 because TMX-2138 also suppressed ASCL1 mRNA levels in the mouse cells (Fig. 6F and fig. S18F). TMX-2172 decreased the half-life of ASCL1 protein (Fig. 6, G and H), consistent with posttranscriptional regulation of ASCL1 by CDK2.

We conducted our screens in IKZF1-independent 293FT cells rather than IKZF1-dependent myeloma cells and in ASCL1-independent Jurkat cells rather than in ASCL1-dependent SCLC cells in an attempt to preserve positive selection. It is possible, however, that some degradation mechanisms will be highly context dependent and restricted to the therapeutic target cell of interest. We also anticipate that some DCK* fusion proteins will not be functional due to steric or conformational effects. This might be remedied by fusing DCK* to the alternative POI terminus (N-terminus versus C-terminus), by exploring different linkers, or using alternative suicide proteins.

IMiDs are important multiple myeloma drugs, but loss of cereblon has emerged as an important mechanism of IMiD resistance (2628). Identification of Spautin-1s mechanism of action could eventually lead to drugs for circumventing this problem.

ASCL1 is a sequence-specific DNA binding transcription factor that would classically be deemed undruggable and serves as a lineage addiction oncoprotein in neural crestderived tumors, such as SCLCs and neuroblastomas (1719, 29). Genetic studies in Xenopus indicate that CDK2 regulates ASCL1 function and that ASCL1 contains multiple potential CDK2 phosphorylation sites that prevent it from inducing neuronal differentiation (30, 31). CDK2 is a potential dependency in some neuroblastomas (3234). CDK2 and N-MYC drive the accumulation of phosphorylated ASCL1 in undifferentiated neuroblastomas (31). Conversely, loss of CDK2 activity, such as through retinoic acidmediated induction of p27 or small-molecule inhibitors, is associated with neuroblastoma differentiation and decreased tumor formation (3237). It will be important to determine how, mechanistically, CDK2 regulates ASCL1 turnover. In particular, we have not yet shown that the regulation of ASCL1 by CDK2 is direct. Nonetheless, our study provides further support for CDK2 as a potential therapeutic target in SCLC and neuroblastoma.

The discovery that the IMiDs reprogram the cereblon ubiquitin E3 ligase for therapeutic benefit has galvanized interest in identifying compounds that can degrade, directly and indirectly, otherwise undruggable proteins. Sometimes, one can engineer heterobifunctional degrader molecules consisting of a POI-binding moiety, a linker, and a ubiquitin-ligase recruitment moiety (38). This approach requires a ligand with suitable binding affinity for the POI, and identifying a successful linker often requires multiple iterations of trial and error. Moreover, this approach fails to harness the many other ways a chemical could directly or indirectly degrade a protein, such as by inhibiting a deubiquitinating enzyme, displacing an interacting protein, or altering protein folding or subcellular localization. A trivial way to down-regulate proteins, especially those with naturally rapid turnovers, is to poison transcription or translation. The screening methodology described here should facilitate the characterization of designer degraders as well as enable mechanism-agnostic searches for compounds and targets that regulate the abundance of previously undruggable proteins.

293FT cells were originally obtained from the American Type Culture Collection (ATCC). 293AD cells were from Cell Biolabs. 293FT CRBN / cells were made by CRISPR-Cas9 editing (see below). 293FT and 293AD cells were maintained in Dulbeccos minimum essential medium (DMEM) supplemented with 10% fetal bovine serum (FBS), penicillin (100 U/ml), and streptomycin (100 g/ml). KMS11, KMS34, MM.1S, and L363 human multiple myeloma cells [gift of K. Anderson (Dana-Farber Cancer Institute)] and Jurkat cells (obtained from ATCC in September 2016) were maintained in RPMI medium supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 g/ml). NCI-H1876 (obtained in November 2016), NCI-H1092 (obtained in November 2018), and NCI-H2081 (obtained in November 2018) were obtained from ATCC. NCI-H1876, NCI-H1092, and NCI-H2081 cells were maintained in DMEM/F12 media supplemented with HITES [10 nM hydrocortisone (Sigma-Aldrich, #H0135), insulin (0.01 mg/ml), human transferrin (0.0055 mg/ml), sodium selenite (0.005 g/ml) (ITS, Gemini, #400-145), and 10 nM -estradiol (Sigma-Aldrich, #E2257)] and 5% FBS. The cell lines 188 and 97-2 were isolated from genetically engineered SCLC mouse tumors (see below for description of cell line generation) and maintained in RPMI 1640 media supplemented with HITES and 10% FBS. All cells were grown at 37C in the presence of 5% CO2. Fresh aliquots of cells were thawed every 4 to 6 months.

The following compounds were purchased: POM (Selleck, #S1567), LEN (Selleck, #S1029), MG132 (N-carbobenzyloxy-l-leucyl-l-leucyl-l-leucinal; Thermo Fisher Scientific, #47479020MG), MLN4924 (Active Biochem, #A-1139), MLN7243 (Thermo Fisher Scientific, #NC1129906), Spautin-1 (BioTechne; #5197/10), cycloheximide (VWR, #97064-724), BVdU (Chem-Impex International Inc., catalog no. 27735), actinomycin D (Thermo Fisher Scientific, #11805017), and dinaciclib (Selleck, #S2768).

CDK2 degraders. Synthesis and characterization of the small-molecule CDK2 degraders TMX-2138 and TMX-2172 and the negative degrader ZXH-7035 (structurally similar to the CDK2 binding region of TMX-2138 and TMX-2172 but lacking the cereblon recruiting element) are described previously (25).

293FT cells stably transduced with bicistronic lentiviruses expressing (i) a fusion between DCK* and the POI and (ii) GFP were seeded at a density of 0.25 106 cells/ml in 25 ml of media in a 15-cm dish (Corning, 353025). Two days later, the cells were counted and resuspended in media to a concentration of 10 106 cells/ml. The sample was passed through a mesh strainer (Thermo Fisher Scientific, #352235). The GFP fluorescence of the cells was analyzed by FACS using a Fortessa Aria II instrument. The brightest 1% of cells were collected in an Eppendorff tube, replated in a six-well dish, and expanded. This process was repeated three to four more times to isolate cells expressing the desired GFP levels.

Jurkat cells were first transduced with PLL3.7-Cas9-IRES-Neo. Neomycin-resistant cells with confirmed Cas9 expression were then superinfected with pLX304-ASCL1-DCK*-IRES-GFP or pLX304-DCK*-IRES-GFP, and transduced cells were selected with blasticidin. The blasticidin-resistant cells were then prepared for FACS sorting as above. In total, the brightest 1% of cells were FACS-sorted three times to isolate cells expressing the desired GFP levels. Jurkat cells expressing Cas9 and DCK*-FOXP3 were made in an analogous manner.

Cell pellets were lysed in a modified EBC lysis buffer [50 mM tris-Cl (pH 8.0), 250 mM NaCl, 0.5% NP-40, and 5 mM EDTA] supplemented with a protease inhibitor cocktail (cOmplete, Roche Applied Science, #11836153001). Whole-cell extracts were quantified using the Bradford protein assay. For experiments with 293FT cells, 10 g of protein per sample was boiled after adding 3 sample buffer (6.7% SDS, 33% glycerol, 300 mM dithiothreitol, and bromophenol blue) to a final concentration of 1; resolved by SDSpolyacrylamide gel electrophoresis (PAGE) using either 12.5% SDS-PAGE, Mini-Protean TGX 4 to 15% gels (Bio-Rad, #456-1086), or Criterion TGX gels (Bio-Rad, #5671085); semi-dry transferred onto nitrocellulose membranes; blocked in 5% milk in tris-buffered saline with 0.1% Tween 20 (TBS-T) for 1 hour; and probed with the indicated primary antibodies overnight at 4C. Membranes were then washed three times in TBS-T, probed with the indicated horseradish peroxidaseconjugated secondary antibodies for 1 hour at room temperature, and washed three times in TBS-T. Bound antibodies were detected with enhanced chemiluminescence Western blotting detection reagents [Immobilon (Thermo Fisher Scientific, #WBKLS0500) or SuperSignal West Pico (Thermo Fisher Scientific, #PI34078)]. The primary antibodies and dilutions used were as follows: rabbit anti-IKZF1 (Cell Signaling Technology, #5443S) at 1:1000, rabbit anti-V5 (Bethyl Laboratories, #A190-120A) at 1:1000, rabbit anti-DCK (Abcam, #151966) at 1:2000, rabbit anti-ASCL1 (Abcam, #ab211327) at 1:1000, rabbit anti-CDK2 (Cell Signaling Technology, #2546S) at 1:1000, mouse anti-P62 (Abcam, #ab56416) at 1:1000, rabbit antiLC3-I and LC3-II (Cell Signaling Technology, #3868S) at 1:1000, rabbit anti-ATG7L (Cell Signaling Technology, #8558S) at 1:1000, rabbit anti-Beclin1 (Cell Signaling Technology, #3495S) at 1:1000, rabbit anti-FIP200 (Cell Signaling Technology, #12436S) at 1:1000, rabbit -phospho-RB1 S795 (Cell Signaling Technology, #9301P) at 1:1000, mouse -RB1 4H1 (Cell Signaling Technology, #9309S) at 1:1000, mouse anti-actin (Sigma-Aldrich; clone AC-15, #A3854) at 1:25,000, mouse anti-Cas9 (Cell Signaling Technology, #14697) at 1:1000, mouse anti-vinculin (Sigma-Aldrich; #V9131) at 1:10,000, and mouse anti-actin (Cell Signaling Technology, #3700S) at 1:10,000. The secondary antibodies and dilutions used were goat anti-mouse (Pierce) at 1:10,000 and goat anti-rabbit (Pierce) at 1:5000.

A total of 750,000 293FT IKZF1-V5 cells per well were seeded in six-well dishes in a volume of 2 ml. On the next day, drugs to be added were diluted from a 10 mM stock (stored at 20C) into 0.5 ml of media before being added to the cells. The final volume in each well was then made up to 3 ml by adding a second drug in 0.5 ml or adding 0.5 ml of drug-free media.

Myeloma cells were seeded in 10-cm plates at a density of 0.75 106 cells/ml in a total volume of 8 ml. On the next day, the desired drug was diluted from a 10 mM stock (stored at 20C) into 1 ml of media, which was added to the intended well to achieve the desired final concentration. The final volume in each well was then made up to 10 ml by adding a second drug in 1 ml or adding 1 ml of drug-free media. After 24 hours, the cells were harvested for analysis.

293FT cells stably transduced with bicistronic lentiviruses encoding (i) IKZF1, DCK*, or DCK*-IKZF1 (IKZF1 cells, DCK* cells, and DCK*-IKZF1 cells, respectively) and (ii) GFP, as well as corresponding EV control cells, were seeded into six-well plates at 20,000 cells per well in 2.5 ml of media. The next day, 1 M BVdU dissolved in DMSO was diluted into media to prepare 6 stock solutions of BVdU at concentrations of 6 mM, 600 M, 60 M, 6 M, and 600 nM. For each stock solution, DMSO concentration was adjusted to a final concentration of 0.6%. Each well in the six-well dish received 0.5 ml of a 6 stock solution of BVdU to achieve final concentrations of 1 mM, 100 M, 10 M, 1 M, and 100 nM, respectively. A total of 0.5 ml of media with 0.6% DMSO was added to the sixth well as a control. Four days later, the cells were collected and counted using a Vi-Cell XR cell counter.

293FT cells were seeded as above at a density of 20,000 cells per well in 2 ml of media. A stock solution of 10 mM POM in DMSO was diluted into media to prepare a 6 M stock solution of POM. Cells received 0.5 ml of the 6 M POM stock solution to achieve an eventual final concentration of 1 M or 0.5 ml of control media. The next day, BVdU was added as described above, and cell proliferation was analyzed as above.

Jurkat cells expressing Cas9 and either ASCL1, ASCL1-DCK*, or DCK* alone were plated at 0.05 106 cells/ml per well in a 12-well plate and treated with increasing concentrations of BVdU (0, 1, 10, 100, 200, or 500 M). Six days later, the cells were counted using a Vi-Cell XR cell counter. For ASCL1 sgRNA rescue experiments, Jurkat cells expressing Cas9 and ASCL1-DCK* cells were superinfected with pLentiGuide-Purobased lentiviruses expressing sgRNAs targeting ASCL1 or a nontargeting sgRNA (sgCTRL). The cells were selected with puromycin, and expression of ASCL1 was analyzed by immunoblot analysis. The cells were then subjected to the BVdU assay as described above.

293FT cells stably transduced with bicistronic lentiviruses encoding (i) DCK*, DCK*-IKZF1, DCK*-K-RAS (G12V), DCK*-Cyclin D1, DCK*-PAX5, DCK*-FOXP3, and DCK*-MYC and (ii) GFP, as well as corresponding EV control cells, were seeded into 384-well plates (Corning, #3764) at 200 cells per well in 30 l of media. The next day, 1 M BVdU dissolved in DMSO was diluted into media to prepare 4 stock solutions of BVdU at concentrations of 4 mM, 2 mM, 1 mM, 400 M, 200 M, 100 M, 40 M, 20 M, 4 M, and 400 nM. On each plate, 10 l of each stock concentration of BVdU was added to two columns (32 wells) to achieve final concentrations of 1 mM, 500 M, 250 M, 100 M, 50 M, 25 M, 10 M, 5 M, 1 M, and 100 nM. Ten microliters of control media was added to four columns. Four days later, the cells were analyzed using an Acumen laser scanning cytometer (TTP Biosciences). GFP fluorescence was quantified by defining the metric GFP-positive object to identify GFP-positive cells while excluding debris or cell fragments.

Determination of Z. DCK* and DCK*-IKZF1 cells were seeded into 384-well plates (Corning, #3764) in 30 l of media at a density of 200 cells per well and allowed to adhere overnight. For each plate, an HPD300 dispenser (Hewlett-Packard) was used to add 4 nl of POM to a final concentration of 1 M to half the wells. An equal volume of DMSO was added to the other half of the plate. The next day, BVdU was added to the entire plate at a concentration of 10 M for the plate of DCK* cells and 100 M for the plate of DCK*-IKZF1 cells. Four days later, the cells were analyzed using an Acumen laser scanning cytometer (TTP Biosciences). The number of GFP-positive objects in each well was measured, and a Z statistic was calculated comparing the POM-treated wells to the DMSO-treated wells.

High-throughput chemical library screening. DCK* and DCK*-IKZF1 293FT cells were seeded into 384-well plates (Corning, #3764) at a density of 200 cells per well in a volume of 30 l of media. A custom-built Seiko Compound Transfer Robot was used to pin transfer 100 nl per well of small-molecule stock solutions from the wells of a drug library plate to the wells of assay plate, such that each well of the assay plate received a unique small molecule. An HPD300 non-contact dispenser (Hewlett-Packard) was used to dispense 100 nl of POM and dipyridamole into columns 23 and 24 and to add 100 nl of DMSO to columns 1 and 2. The final concentrations of POM and dipyridamole were 10 M and 12.5 M, respectively. The next day, 10 l of BVdU stock solution was added to columns 2 to 24 of each of the DCK-IKZF1 and DCK* assay plates, respectively. The concentration of the BVdU stock solution was calculated to achieve the desired final concentration of BVdU (10 M in DCK* assay plates and 100 M in DCK*-IKZF1 assay plates) in the well.

After 4 days, the GFP fluorescence of each assay plates was quantified using an Acumen scanning laser cytometer. For each plate, the average and SD of the GFP fluorescence of wells in columns 3 to 22 were calculated. For each well on an assay plate, the GFP fluorescence was converted to a z score using the formula: z(well) = [GFP (well) GFP (plate)] / GFP (plate), where GFP (plate) is the mean GFP fluorescence for that plate and GFP (plate) is the SD for that plate.

High-throughput chemical library screening (in-well competition assay). DCK*-IKZF1 (GFP) and DCK* (Td) cells were mixed together in a 1:1 ratio and then seeded into 384-well plates (Corning, #3764) at a density of 400 cells per well in 30 l of media. Pin transfer from IMiD derivative library plates and dispensation of POM and dipyridamole were performed as described above. The next day, 10 l of BVdU stock solution was added to columns 2 to 24 of each plate to achieve a final concentration of 100 M. After 4 days, the GFP and TdTomato fluorescence of each assay plate was quantified using an Acumen scanning laser cytometer. For each well, the ratio of GFP/tdTomato fluorescence was calculated and normalized to the values in the well that received DMSO and BVdU. The resulting values were converted to a heatmap using Morpheus (Broad Institute).

Determination of Z. 293FT IKZF1-Fluc cells were seeded into 96-well plates at a density of 2000 cells per well in a volume of 50 l of media and incubated overnight at 37C. The next day, an additional 50-l media and POM (final concentration of 2 M) was added to 30 wells of the plate (rows B to G, columns 2 to 6). Control media containing DMSO was added to 30 wells of the plate (rows B to G, columns 7 to 11). A Dual-Glo assay (Promega) was performed by first aspirating all media from the tissue culture plates. Twenty-five microliters of a 1:1 dilution of Dual-Glo luciferase assay reagent in phosphate-buffered saline (PBS) was added to wells and incubated for 10 min. Luminescent signal was measured with a plate reader. Stop & Glo reagent (12.5 l) was then added to the wells, incubated for 10 min, and luminescent signal was measured. The average Fluc/Rluc ratio for cells treated with DMSO and POM was calculated, and a Z statistic was calculated.

High-throughput library screening using Fluc/Rluc readout. IKZF1-Fluc assay plates were generated by plating 293FT IKZF1-Fluc cells into 384-well plates. For the 8-hour treatment arm, cells were plated at a density of 4000 cells per well. A custom-built Seiko Compound Transfer Robot was used to pin transfer 100 nl per well of small molecule from the drug library plate to the assay plate, such that each well of the assay plates received a unique small molecule. After 8 hours, the plates were shaken out and blotted on clean paper towels to remove the media. A Thermo Multidrop Combi was used to dispense 20 l of a 1:1 dilution of Dual-Glo luciferase reagent, and the plates were shaken for 10 min. Firefly luciferase signal was quantified using an EnVision plate reader. A Thermo Multidrop Combi was used to dispense 10 l of Dual-Glo Stop + Glo reagent, and the plates were shaken for 10 min. Renilla luciferase signal was quantified using an EnVision plate reader. For each plate, the ratios of the Firefly/Renilla luciferase signals were converted to a Z-distribution as outlined above. For the 4-day treatment arm, the experiment was performed in an analogous manner, but the cells were plated at a density of 200 cells per well and were incubated for 4 days before analysis.

Gene-targeting sgRNAs and appropriate controls were designed using the rule set described at the Genetic Perturbation Program (GPP) portal (http://portals.broadinstitute.org/gpp/public). Oligonucleotides were flanked by polymerase chain reaction (PCR) primer sites, and PCR was used to amplify DNA using NEBNext kits. The PCR products were purified using Qiagen PCR cleanup kits and cloned into pXPR_BRD003 using Golden Gate cloning reactions. Pooled libraries were amplified using electrocompetent Stbl4 cells. Viruses were generated as outlined at the GPP portal. The sgRNA library (CP1080, M-AB34) was custom-designed to target cancer-relevant druggable genes. It consisted of 5566 sgRNAs targeting 788 genes (7 sgRNAs targeting each gene) and 300 nontargeting sgRNAs as controls (table S6).

Jurkat cells that had been infected with PLL3.7-Cas9-IRES-Neo and subsequently maintained in G418 were then superinfected with pLX304 ASCL1-DCK*-V5-IRES-GFP or pLX304 DCK*-V5-IRES-GFP and placed under blasticidin selection. Blasticidin-resistant cells were sorted for GFP expression (top 1%) three times by FACS. Protein abundance of ASCL1-DCK* or DCK* alone was confirmed by immunoblot analysis, and functionality of ASCL1-DCK* or DCK* alone was determined using BVdU sensitivity and rescue experiments with sgRNAs targeting ASCL1. Cas9 expression was confirmed by immunoblot analysis, and Cas9 activity was confirmed using a Cas9 GFP reporter [pXPR_011 (Addgene, #59702)] (39) that showed near maximal editing 10 days after infection.

On day 0, ASCL1-DCK* and DCK* cells expressing Cas9 were expanded and then counted. For each line, 2.2 107 cells (4000 cells per sgRNA) were pelleted and resuspended at 2 106 cells/ml in media supplemented with polybrene (8 g/ml) and infected at a multiplicity of infection (MOI) of ~0.3 with the sgRNA druggable library (CP1080, M-AB34) described above. The cells mixed with polybrene and virus were then plated in 1-ml aliquots onto 12-well plates and centrifuged at 434g for 2 hours at 30C. Sixteen hours later (day 1), the cells were collected, pooled, and centrifuged to remove the virus and polybrene, and the cell pellet was resuspended in complete media at 2 105 cells/ml and plated into nontissue culturetreated t175 flasks. The cells were then cultured for 48 hours before being placed under puromycin (1 g/ml) drug selection at 4 105 cells/ml.

A parallel experiment was performed on day 3 to determine the MOI. To do this, the cells infected with the sgRNA library and mock-infected cells were plated at 4 105 cells/ml in the presence or absence of puromycin. After 72 hours (day 6), cells were counted using the Vi-Cell XR cell counter, and the MOI was calculated (which ranged from 0.2 to 0.3 for each replicate) using the following equation: (# of puromycin-resistant cells infected with the sgRNA library / # total cells surviving without puromycin after infection with the sgRNA library) (# of puromycin-resistant mock-infected cells / # total mock-infected cells).

On day 6 after MOI determination, puromycin-resistant cells were pooled, collected, and counted, and 1 108 cells were replated at a concentration of 4 105 cells/ml in complete media containing puromycin (1 g/ml). The remaining cells were discarded. On day 8, again, the puromycin-resistant cells were pooled, collected, and counted, and 1 108 cells were replated at a concentration of 4 105 cells/ml in complete media containing puromycin (1 g/ml).

On day 10, puromycin-resistant cells were pooled, collected, and counted. A total of 2 107 cells were collected and washed in PBS, and the cell pellets were frozen for genomic DNA isolation for the initial time point before BVdU selection. Then, 2 107 cells were resuspended in complete media (now without puromycin) containing either 200 or 500 M BVdU at a final concentration of 5 104 cells/ml and plated into t175-cm flasks. Thus, at least 1000 cells per sgRNA were introduced into BVdU selection.

On day 15, cells treated with 200 or 500 M BVdU were collected and counted. A total of 10 106 cells from each arm of the screen were then resuspended in complete media containing either 200 or 500 M BVdU at a final concentration of 5 104 cells/ml and plated into t175-cm flasks. The remaining cells were centrifuged and washed in PBS, and the cell pellets were frozen. Again, at least 1000 cells per sgRNA were maintained under BVdU selection.

On day 20, cells treated with 200 or 500 M BVdU were collected and counted. A total of 10 106 cells from each arm of the screen were then resuspended in complete media containing either 200 or 500 M BVdU at a final concentration of 5 104 cells/ml and plated into t175-cm flasks. If available, the remaining cells were centrifuged and washed in PBS, and the cell pellets were frozen. Again, at least 1000 cells per sgRNA were maintained under BVdU selection.

On day 25, all remaining cells were collected and counted. The remaining cells were divided in aliquots of 6 106 cells (which corresponds to 1000 cells per sgRNA) and washed in PBS, and the cell pellets were frozen for genomic DNA isolation for the final time point after BVdU selection. The screen was performed in two biological replicates.

Following completion of the screen, genomic DNA was isolated using a Qiagen Genomic DNA midi prep kit (catalog no. 51185) according to the manufacturers protocol. Raw Illumina reads were normalized between samples using log2[(sgRNA reads/total reads for sample) 1 106 + 1]. The initial time point data (day 10) were then subtracted from the end time point after BVdU selection (day 25) to determine the relative enrichment of each individual sgRNA after BVdU treatment using hypergeometric analysis and the STARS algorithm. A q value cutoff of <0.25 was used to call hits. The averaged data from two biological replicates were used for all analyses.

293FT cells stably transduced with bicistronic lentiviruses encoding (i) DCK*-IKZF1 and GFP and (ii) DCK* and TdTomato were mixed together at a ratio of 1:99. Pooled cells were plated at a density of 20,000 cells per well of a six-well plate and in a total volume of 2 ml of media. Cells received 0.5 ml of the 6 M POM stock solution to achieve an eventual final concentration of 1 M or 0.5 ml of control media. The next day, the cells received 0.5 ml of 600 M BVdU stock solution or 0.5 ml of control media. Cells were collected for FACS analysis on days 0, 3, 6, 10, and 14. After each time point, cells were reseeded at 20,000 cells per well and treated with fresh BVdU (or DMSO).

DCK*-IKZF1 293FT cells were infected with a mixture of two lentiviruses encoding Cas9 and either (i) sgDCK and mCherry or (ii) sgCTRL and BFP (blue fluorescent protein). The two lentiviruses were mixed together such that the ratio of mCherry-positive to BFP-positive cells after infection and puromycin selection was 1:99. An analogous experiment was set up using a lentivirus encoding sgCTRL and mCherry. The pool of infected cells was plated at 20,000 cells per well in a six-well plate and then cultured in media containing either 100 M BVdU or DMSO for 21 days. Cells were collected for FACS analysis on days 1, 6, 18, and 33. After each time point, cells were reseeded at 20,000 cells per well and treated with fresh BVdU.

Jurkat cells that had been stably infected to express Cas9 and DCK*-FOXP3 were superinfected with lentivirus encoding either (i) sgDCK and mCherry or (ii) sgCTRL and BFP. These cells were mixed together and analyzed by FACS to achieve a final ratio of mCherry-positive to BFP-positive cells of 1:99. The pool of infected cells was plated at 40,000 cells/ml in a six-well plate and then cultured in media containing either 100 M BVdU or DMSO for 14 days. Cells were collected for FACS analysis on days 6 and 14.

The Jurkat cells expressing Cas9 and ASCL1-DCK* that were used for the CRISPR-Cas9 screen described above were superinfected with lentiviruses encoding sgRNAs targeting CDK2, ASCL1, or a nontargeting sgRNA as a control, the fluorescent protein mCherry and a puromycin resistance gene or with a lentivirus encoding a nontargeting sgRNA as a control, and the fluorescent protein BFP and a puromycin resistance gene (see schema in fig. S16F). The cells were selected with puromycin. mCherry puromycin-resistant cells were then mixed with BFP puromycin-resistant cells at a 1:3 ratio as determined by FACS analysis. The mixed cells were plated at 5 104 cells/ml and then cultured in media containing 500 M BVdU or DMSO (0) for 18 days. FACS analysis was performed every 6 days. After each FACS analysis, fresh BVdU was added, and the density of the cells was adjusted to 5 104 cells/ml with fresh media.

Cells were counted using a Vi-Cell XR cell counter and were plated at a concentration of 4 105 cells/ml per well for NCI-H1092, 188, and 97-2 SCLC cell lines or at 1 106 cells/ml per well for the NCI-H1876 SCLC cell line in six-well plates. Cells were then treated with the CDK2 degraders TMX-2138, TMX-2172, or the negative degrader ZXH-7035 (Neg Deg) at 500 nM for 36 hours for NCI-H1092 and NCI-H1876 human SCLC cell lines or 8 hours for 188 and 97-2 mouse SCLC cell lines. For half-life time determination with cycloheximide, 97-2 cells were treated with CDK2 degraders for 4 hours before the addition of cycloheximide at 150 g/ml. Cells were harvested at the indicated times after addition of cycloheximide.

293FT cells were seeded in six-well plates at a density of 750,000 cells per well in 2.5 ml of media per well. The next day, the cells were treated with the indicated concentrations of Spautin-1, POM, or DMSO for 24 hours. Multiple myeloma cells were seeded in 10-cm plates at a density of 0.75 106 cells/ml in a total volume of 9 ml of media. The next day, the cells were treated with the indicated concentrations of Spautin-1, POM, or DMSO. RNA was extracted using an RNeasy mini kit (Qiagen, #74106) according to the manufacturers instructions. RNA concentration was determined using the NanoDrop 8000 (Thermo Fisher Scientific). cDNA was generated by reverse transcription using the AffinityScript qPCR (quantitative PCR) cDNA Synthesis kit (Agilent, 600559) according to the manufacturers instructions. qPCR was performed using the LightCycler 480 (Roche) with the LightCycler 480 Probes Master Kit (Roche) and TaqMan probes (Thermo Fisher Scientific) according to the manufacturers instructions. The Ct values for each probe were then normalized to the Ct value of ACTB for that sample. The data from each experiment were then normalized to the control to determine the relative fold change in mRNA expression. The following TaqMan probes were used: Hs00958474_m1 (IKZF1 human), ASCL1 human Hs04187546_g1 for detection of endogenous ASCL1, ASCL1 human Hs05000540_s1 for detection of the exogenous ASCL1-DCK* fusion, ACTB human Hs01060665_m1, Ascl1 mouse Mm03058063_m1, and Actb mouse Mm00607939_s1. All quantitative calculations were performed using the 2Ct method using Beta Actin (ACTB) as a reference gene.

For the positive selection small-molecule screen, GFP fluorescence for each well was normalized to untreated wells. For each library drug, normalized GFP fluorescence was plotted as a function of library drug concentration. Each drug treatment was performed in duplicate. Data were analyzed and plotted using GraphPad Prism v6, median inhibitory concentration (IC50) values were determined using the log (inhibitor) versus response -- Variable slope (four parameters) analysis module, and area under the curve (AUC) values were determined using the AUC analysis module (13). For the positive selection CRISPR-Cas9 BVdU resistance screen, the relative fold enrichment of each individual sgRNA after BVdU treatment was calculated using both Broad Institutes hypergeometric analysis and the STARS algorithm to determine a rank list of candidate ASCL1 stabilizer genes ranked by q value, where statistical significance is q < 0.25.

For all other experiments, statistical significance was calculated using unpaired, two-tailed Students t test. P values were considered statistically significant if the P value was <0.05. For all figures, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Error bars represent SD unless otherwise indicated.

Acknowledgments: We thank the members of the Kaelin and Oser laboratories for helpful discussions. Special thanks to the ICCB (Longwood) at HMS for assistance with small-molecule screens, to W. Gao for generation of adenoviral vectors used for recombination cloning, and D. Hong for generation of SCLC mouse cell lines. Funding: W.G.K. is supported by an NIH R35 grant and is an HHMI investigator. M.G.O. is supported by a Damon Runyon Cancer Research Foundation Clinical Investigator Award and an NCI/NIH KO8 grant (no. K08CA222657). V.K. is supported by an American Society of Hematology Research Training Award and T32 NIH Training Grant CA009172. J.A.P. is funded by an NIGMS grant R01 GM132129. E.S.F. is funded by NCI R01CA2144608. N.S.G. is funded by NIH R01 CA214608-03. M.I. is supported by an Internationalisation Fellowship from the Carlsberg Foundation. C.J.O. is supported by an NIH/NCI Pathway to Independence Award (R00CA190861). Author contributions: V.K., L.D., and B.L.L. performed experiments and, together with W.G.K. and M.G.O., designed experiments, analyzed data, and assembled and wrote the manuscript. J.A.M. helped design experiments. A.C.W. and A.H.S. performed experiments. M.I. designed and synthesized the IMiD library. J.P. and C.J.O. measured CRBN binding and cellular activity of candidate IMiDs; J.B. supervised these experiments. I.S.H. and J.E.E. constructed the Ludwig anticancer and antimetabolite libraries and helped analyze data from the screen. E.D., X.L., and S.J.B. synthesized and characterized Spautin-1 derivatives. J.A.P. performed TMT global proteomic profiling of Spautin-1. S.P.G. supervised these experiments. K.A.D. and E.S.F. analyzed TMT proteomic data. K.J.B. determined the half-lives of luciferase fusion proteins and the Z of the dual-luciferase system. J.G.D. helped analyze data from the CRISPR screen. M.T., T.Z., and N.S.G. helped generate and validate CDK2 degraders. Competing interests: W.G.K. has financial interests in Lilly Pharmaceuticals, Fibrogen, Agios Pharmaceuticals, Cedilla Therapeutics, Nextech Invest, Tango Therapeutics, and Tracon Pharmaceuticals. N.S.G. is a founder, science advisory board member, and equity holder in Gatekeeper, Syros, Petra, C4, B2S, Aduro, and Soltego (board member). E.S.F. is a founder, scientific advisory board (SAB) member, and equity holder of Civetta Therapeutics, Jengu Therapeutics (board member), and Neomorph Inc. E.S.F. is an equity holder of C4 Therapeutics. E.S.F. consults or has consulted for Novartis, AbbVie, Astellas, Deerfield, EcoR1, and Pfizer. The Fischer laboratory receives or has received research funding from Novartis, Deerfield, and Astellas. The Gray laboratory receives or has received research funding from Novartis, Takeda, Astellas, Taiho, Janssen, Kinogen, Voronoi, Her2llc, Deerfield, and Sanofi. M.G.O. has sponsored research agreements with Lilly Pharmaceuticals and Takeda Pharmaceuticals. V.K. has consulted for Cedilla Therapeutics. S.J.B. is on the SAB of Adenoid Cystic Carcinoma Foundation. J.B. is an employee, executive, and shareholder of Novartis AG (Basel, Switzerland). J.G.D. consults for Agios, Foghorn Therapeutics, Maze Therapeutics, Merck, and Pfizer; J.G.D. consults for and has equity in Tango Therapeutics. J.G.D.s interests were reviewed and are managed by the Broad Institute in accordance with its conflict of interest policies. I.S.H. is a consultant for ONO Pharmaceuticals (USA). V.K. and W.G.K. are inventors on a patent application on positive selection assays to identify protein degraders, which was filed by the Dana-Farber Cancer Institute (U.S. patent application number 16/332,921, filed on 13 March 2019 and published on 1 August 2019). N.S.G., M.T., and T.Z. are named inventors on patent applications covering Cdlk2 degraders described in the paper, and which were filed by the Dana Farber Cancer Institute (U.S. Provisional Application No. 62/829,302, filed April 4, 2019 and U.S. Provisional Application No: 62/981,334, filed February 25, 2020). The authors declare that they have no other competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. All plasmids are available from the authors.

See the article here:
Targeting oncoproteins with a positive selection assay for protein degraders - Science Advances

The Computer Weekly Developer Network examines the impact of Covid-19 (Coronavirus) on the software application development community.

With only a proportion of developers classified as key workers (where their responsibilities perhaps included the operations-side of keeping mission-critical and life-critical systems up and online), the majority of programmers will have been forced to work remotely, often in solitude.

So how have the fallout effects of this played out?

This post comes from Justin Reock in his role as chief evangelist for open source software (OSS) & Application Programming Interface (API) management at Perforce Software.

Reock reflects upon the use of open source platforms, languages and related technologie in general in light of the Covid-19 global crisis and writes as follows

On the whole, I would argue that open source software has been invaluable during the pandemic.

Crowd-sourced software initiatives and hackathons, protein-folding peer-to-peer networks and foundation sponsorship have all been in play throughout the contagion and many of these initiatives continue forwards.

GitHub has shown us that commits held steady or even increased suggesting (if it is fair to measure that in terms of raw commits without considering quality) that developer productivity has held steady or even gone up.

For many developers, having a shared project and sense of community during a very isolating time for humanity has been uplifting and good for their spirits. Its a reminder that coding together is in fact a social activity, no different than any other collaborative and creative endeavour.

Perhaps the biggest impact and fallout from this whole period of experiences (for programmers, operations staff and the wider software engineering community) will be the acceleration of transformation and DevOps initiatives within businesses.

So many have witnessed the resilience of businesses that have already undergone the DevOps transition (and even watched their profits soar) as we moved to online ordering, contactless delivery and more.

The CI/CD part of the DevOps makeover has always been about dealing with constant change.The mantra of releases are hard, so release often embraces the notion that change is difficult, so organisations should make themselves really good at dealing with it. That meant when the pandemic hit, the seams of our global digital twin were tested. Companies that were capable of quickly refactoring to online experiences, digital goods and other conveniences have now become essential to carrying on a reasonable quality of life in the physical world.

It is one thing to expect the unexpected, and it is quite another to design systems that thrive in unexpected conditions.Whatever requisite effort may need to be invested to achieve DevOps maturity in an organisation, the positive impact it can have to business longevity is now indisputable.

However, especially in segments of the industry that are highly collaborative such as gaming, quality and deadlines have suffered drastically and development teams have blamed it squarely on moving to a remote work model.

NOTE: As a software change management specialist, Perforce has a particularly acute proximity with and close understanding of how games programmers work.

Even enabling employees to work from home was a challenge, as the hardware supply chain which we rely on to deliver our webcams, tablets, and laptops and other tech gear suffered major disruptions: so, all in all, there is no question that organisations, including open source communities, which had already taken steps towards transformation and remote work were able to continue operations smoothly, though not completely without impact.

That said, the overall industry picture is not all rosy, with many segments that rely heavily on peer collaboration taking a hit in quality and productivity.

We hope, of course, for brighter future times for all.

Reock: Commit to commit dear developers, you know you want to.

Go here to see the original:
Programming in the pandemic - Perforce: In open source, crowd is a positive - ComputerWeekly.com

Sean McClain, Co-Founder and CEO of AbSci.

There has been a lot of recent attention on the challenges of delivering COVID-19 vaccines. But there are also challenges in making them. For some of the newer options like those from Johnson & Johnson and Oxford-AstraZeneca, the modified cells used in vaccine production are struggling under the scale of demand. But synthetic biology company AbScis recent acquisition of the artificial intelligence platform, Denovium, could help mitigate this type of challenge in the future.

Unlike mRNA vaccines, the Johnson & Johnson/Oxford-AstraZeneca class of vaccines rely on a type of virus called adenovirus which is known to cause colds in chimpanzees. To address COVID-19, the adenovirus is genetically altered to express the SARS-CoV-2 spike protein which is what ultimately triggers the bodys immune response. Like mRNA vaccines, adenovirus-based vaccines train the body to recognize and fight COVID-19, foregoing the need to inject a person with a weakened version of SARS-CoV-2.

But producing enough adenovirus cells has been a challenge. To make vaccine doses, large volumes of altered adenovirus are produced by replicating cells in bioreactors. But, the scale of production can also cause the cells to weaken. This can result in a reduced output of adenovirus copies. So while these new vaccines may represent a breakthrough in adenovirus-based therapeutics, the process also highlights some critical roadblocks.

One major issue is that drug discovery and drug manufacturing are often disconnected from one another. Drug discovery typically starts with screeningthe process of finding a set of compounds out of 100,000 combinations that can best neutralize a targeted weak point of a disease. But when a promising protein is identified, it often turns out to be difficult to scale effectively.

Once a therapeutic compound is identified, researchers must then determine if it works well with a group of similar cells called a cell-line. By inserting the compound into the cellswhich then divide and multiply in a bioreactorthe cells act like factories to produce greater volumes of the compound of choice. But, as in the case with adenovirus-producing cells, not all cells can maintain their functions at large volumes. If the protein compound doesnt work well in a scalable cell-line, researchers often have to go back to the drawing board to find a new compound and start again.

Many in the biopharma space are aware of this inefficient process. The synthetic biology company AbSci has spent years developing a platform solution that streamlines the workflow. [Our platform] is simultaneously a drug discovery and manufacturing platform that allows you to discover your drug and the cell line that can manufacture [it], says AbSci CEO, Sean McClain. Were finally uniting drug discovery and manufacturing the first time.

AbSci refers to their core process as their Protein Printing platform, not because it uses ink and paper to make proteins but as an analogy for ease and speed. The first technology [in our platform] is our SoluPro E. coli strain. It has been highly engineered to be more mammalian-like to be able to produce mammalian-like proteins that E. coli wasn't previously capable of doing, says McClain. AbSci also uses what the company calls a folding solution to precisely tailor how proteins fold and therefore function.

Imad Ajjawi, Co-Founder and CBO of Denovium

To find the most effective protein, AbSci alters its folding solutions to create as many protein varieties as possible, often to the order of 10s of millions. The more protein types available, which AbSci refers to as libraries, the higher the likelihood of success. But this also creates a challenge: so many options, but which to choose?

To address this, AbSci recently acquired artificial intelligence company, Denovium. By integrating Denoviums AI platform, AbSci can improve its data analysis via AI models. From there, the company can take the best candidates and find the most effective cell-line to produce the chosen compounds at scale. McClain explains that traditional drug discovery and manufacturing typically takes years. But AbScis platform can take that timeline down to weeks. Were actually able to manufacture [therapeutics] because the dirty secret in pharma is that so many drugs get shelved because [pharma companies] can't actually manufacture them, says McClain.

For McClain, acquiring Denovium is a big step forward for AbScis discovery process. Its going to change the paradigm. Its really a perfect marriage of both data and AI technology. If you don't have good data feeding into your AI model, it's worthless. But if you don't have an AI technology, you can't mine [the data] and get all the benefits, says McClain.

Denoviums co-founder and CBO, Imad Ajjawi, also sees the new collaboration as a significant opportunity. It's really exciting to be a part of AbSci because they have all the data, billions of points that the deep learning engine can now analyze, says Ajjawi. AbScis acquisition also comes on the heels of the companys $65 million Series E in late 2020.

Upgrading the union of biology and AI is important for advancing synthetic biology innovation. But the true potential beneficiaries of this advanced discovery platform are those in need of novel drug options.

AbScis main goal as a company is to bring therapeutics to market more quickly. This technology's impact on healthcare is profound because more drugs and biologics can now enter patients' hands faster, says McClain.

McClain believes that AbScis technology will help speed the process of clinically testing new medications. Faster clinical trial turnarounds could increase the number of drugs approved to address a range of diseases. This could be most impactful for patients with rare or difficult to treat conditions as drug discovery is often prioritized based on how long it takes to find a scalable cell-line.

But though AbSci is working to accelerate drug discovery, the process still takes time. Right now, we have six drugs that are in preclinical or clinical trials. And one of them is actually in phase three. So we could have an improved product here in the next couple of years, says McClain.

As Absci and Denovium finalize their technology integrations, McClain is also looking ahead to build as many partnerships as possible. The more partnerships we do, the more patients were able to affect that at the end of the day, says McClain.

In line with that goal, AbSci today announced a continuation of its partnership with Astellas and Xyphos. AbSci will take on screening and identifying an optimal cell-line for a leading variant of Xyphos MicAbody, a bispecific antibody-like adaptor molecule used in the company's immuno-oncology program.

McClain expects more partnership announcements will follow in the first quarter of 2021. We have some really exciting partnerships that are going to be coming out over this next quarter that I think speak to the [range] of the types of disease states we're working on and the breadth of how the technology can be used within biopharma, says McClain.

Im the founder of SynBioBeta, and some of the companies that I write about are sponsors of the SynBioBeta conference and weekly digest, including AbSci. Thank you to Fiona Mischel and Vinit Parekh for additional research and reporting in this article.

Read more:
Synthetic Biology Startup Acquires AI Platform To Disrupt The Drug Industry - Forbes

BEDMINSTER, N.J.--(BUSINESS WIRE)--Tyme Technologies, Inc. (NASDAQ: TYME), an emerging biotechnology company developing cancer metabolism-based therapies (CMBTs), announced that it has received notification that the United States Patent and Trademark Office has granted additional patent claims related to the Companys metabolomic technology platform. The patent, U.S. Patent No. 10,905,698, is directed to methods for treating COVID-19.

Unlike immune therapies that depend upon the structure of the external virus coat of COVID-19 where the therapy directs its attack, we believe TYME-19 is agnostic to this structure and any mutations to the viral coat. Like other TYME agents, TYME-19 affects cellular metabolism. It constrains viral replication after a virus has inserted its genetic blueprint into an infected cell by inhibiting the ability of the virus to use the cells synthetic apparatus to make viral proteins and lipids. As a result, we believe that TYME-19 diminishes the ability of COVID-19 to hijack an infected cell. TYME intends to initiate the appropriate clinical trials to substantiate the safety and efficacy of TYME-19.

TYME-19 is an investigational compound that is not approved in the U.S. for any disease indication.

About TYME-19

TYME-19 is an oral synthetic member of the bile acid family that the Company also uses in its anticancer compound, TYME-18. Because of its expertise in metabolic therapies, the Company was able to identify TYME-19 as a potent, well characterized antiviral bile acid and has performed preclinical experiments establishing effectiveness against COVID-19. Bile acids have primarily been used for liver disease; however, like all steroids, they are messenger molecules that modulate a number of diverse critical cellular regulators. Bile acids modulate lipid and glucose metabolism and can remediate dysregulated protein folding, with potentially therapeutic effects on cardiovascular, neurologic, immune, and other metabolic systems. Some agents in this class also have antiviral properties. In preclinical testing, TYME-19 repeatedly prevented COVID-19 viral replication without attributable cytotoxicity to the treated cells. Previous preclinical research has also shown select bile acids like TYME-19 have had broad antiviral activity.

About Tyme Technologies

Tyme Technologies, Inc., is an emerging biotechnology company developing cancer therapeutics that are intended to be broadly effective across tumor types and have low toxicity profiles. Unlike targeted therapies that attempt to regulate specific mutations within cancer, the Companys therapeutic approach is designed to take advantage of a cancer cells innate metabolic weaknesses to compromise its defenses, leading to cell death through oxidative stress and exposure to the bodys natural immune system.

With the development of TYME-18 and TYME-19, the Company believes that it is also emerging as a leader in the development of bile acids as potential therapies for cancer and COVID-19. For more information, visit http://www.tymeinc.com. Follow us on social media: Facebook, LinkedIn, Twitter, YouTube and Instagram.

Forward-Looking Statements/Disclosure Notice

In addition to historical information, this press release contains forward-looking statements under the Private Securities Litigation Reform Act that involve substantial risks and uncertainties. Such forward-looking statements within this press release include, without limitation, statements regarding our drug candidates (including SM-88 and TYME- 18) and their clinical potential and non-toxic safety profiles, our drug development plans and strategies, ongoing and planned preclinical or clinical trials, including the proposed TYME-19 proof-of-concept study, preliminary data results and the therapeutic design and mechanisms of our drug candidates. The words believes, expects, hopes, may, will, plan, intends, estimates, could, should, would, continue, seeks, anticipates, and similar expressions (including their use in the negative) are intended to identify forward-looking statements. Forward-looking statements can also be identified by discussions of future matters such as: the effect of the novel coronavirus (COVID-19) pandemic and the associated economic downturn and impacts on the Company's ongoing clinical trials and ability to analyze data from those trials; the cost of development and potential commercialization of our lead drug candidate and of other new products; expected releases of interim or final data from our clinical trials; possible collaborations; and the timing, scope, status, objectives and strategy of our ongoing and planned trials; the success of management transitions; and other statements that are not historical. The forward-looking statements contained in this press release are based on managements current expectations and projections which are subject to uncertainty, risks and changes in circumstances that are difficult to predict and many of which are outside of our control. These statements involve known and unknown risks, uncertainties and other factors which may cause the Companys actual results, performance or achievements to be materially different from any historical results and future results, performance or achievements expressed or implied by the forward-looking statements. These risks and uncertainties include but are not limited to: the severity, duration, and economic and operational impact of the COVID-19 pandemic; that the information is of a preliminary nature and may be subject to change; uncertainties inherent in the cost and outcomes of research and development, including the cost and availability of acceptable-quality clinical supply, and in the ability to achieve adequate start and completion dates, as well as uncertainties in clinical trial design and patient enrollment, dropout or discontinuation rates; the possibility of unfavorable study results, including unfavorable new clinical data and additional analyses of existing data; risks associated with early, initial data, including the risk that the final data from any clinical trials may differ from prior or preliminary study data; final results of additional clinical trials that may be different from the preliminary data analysis and may not support further clinical development; that past reported data are not necessarily predictive of future patient or clinical data outcomes; whether and when any applications or other submissions for SM-88 may be filed with regulatory authorities; whether and when regulatory authorities may approve any applications or submissions; decisions by regulatory authorities regarding labeling and other matters that could affect commercial availability of SM-88; the ability of TYME and its collaborators to develop and realize collaborative synergies; competitive developments; and the factors described in the section captioned Risk Factors of TYMEs Annual Report on Form 10-K filed with the U.S. Securities and Exchange Commission on May 22, 2020, as well as subsequent reports we file from time to time with the U.S. Securities and Exchange Commission available at http://www.sec.gov.

The information contained in this press release is as of its release date and TYME assumes no obligation to update forward-looking statements contained in this release as a result of future events or developments.

Originally posted here:
TYME Granted U.S. Patent Claims Covering Use of TYME-19 to Treat COVID-19 Infections - Business Wire

Data science, machine learning, and AI experts highlight the top AI and ML trends they expect to shape the data science industry in 2021

The ongoing impact of Covid-19 is still affecting organizations nearly a year since the pandemic began, with business leaders continuing to leverage technology in order to navigate the crisis. According to Oxylabs dedicated AI and ML advisory board, some of the most important trends in 2021 will include the increased use of ethical AI for diversity, accountability, and model explainability, alongside increased instances of AI solving challenging real-world problems.

Oxylabs advisory board comprises the leading figures in the machine learning, AI, and data science industries and its members outline what they believe are the most important data science predictions for the year ahead:

Firstly, Pujaa Rajan, Machine Learning Engineer at Stripe, USA Ambassador at Women in AI andGoogleDeveloper MLExpert, believes COVID-19 will instigate a renewed enthusiasm for the application of edge AI in the healthcare industry and the use of ethical AI:

Covid-19 defined 2020 and although development in healthcare has historically been slower than other industries due to regulation this year will see a focus on edge AI in the healthcare industry and other industries. This will lead to the ability to run ML models locally, and tiny ML, resulting in smaller sized ML models that fit on smaller devices like phones. Businesses will focus on these specific, technical areas because they are related to data privacy and security, which the general public and government increasingly care about.Model explainability and interpretability is a space that the government, healthcare companies and finance companies are all actively exploring because of technical curiosity and business motivations. Many leaders will also finally prioritise AI ethics, diversity, inclusion, model explainability, and model interpretability after public outrage at many bad, biased, and unethical applications of AI. On the other hand, the biggest AI news last year was OpenAIs GPT-3, so I expect continued innovation in large NLP models. Software and hardware are like yin and yang. Since the larger models will need more efficient hardware, neural network accelerators will be a hot space.

Ali Chaudhry, PhD researcher, Artificial Intelligence atUCL, sees AI as having have more of a contribution in solving challenging real-world problems in 2021:

I think there will be more focus on fairness, transparency, accountability and explainability in AI systems this year, hence, we can expect more regulations from governments around the globe. We will also see AIs contribution in solving more challenging real-world problems, similar to the protein folding problem that was recently solved by AI. In terms of AI techniques that are set to emerge, there will be more real-world applications of Reinforcement Learning (RL) algorithms and RL will also retain its top position in academia.

Another prediction comes from Gautam Kedia, Machine Learning Engineering Manager at Stripe, ex-Applied Scientist Lead at Microsoft, previously Head of Applied ML at Lyft. He considers how AI-generated content could finally become mainstream across multiple sectors:

AI-generated content will become mainstream and in the next few years, I expect truly generative models to be producing logos, short stories, stock images, voiceovers and workouts, DALL-E is just a start and I believe this content will gradually start to pass the Turing Test. Self-driving cars will also take another step forward and I expect Waymo to start a taxi service directly competing with Uber & Lyft. Tesla will also release the much-awaited Full Self Driving computer.

Finally, Jonas Kubilius, AI researcher, Marie Skodowska-Curie Alumnus, and Co-Founder of Three Thirds is optimistic about the implementation of AI in healthcare but also has fears that AI investment may suffer:

Im certainly optimistic about AI-driven solutions making a greater impact in the healthcare sector and drug discovery, however, my only concern is the economic impact of the global COVID-19 pandemic. It may well be that there is a slowdown of investments in AI-driven solutions and research labs, forcing companies to justify any investments they make and focus very clearly on problems where AI brings a clear added value. With ever increasing pressure on governments and organisations to take action in regard to climate change, I expect to see more AI-driven solutions being leveraged in this field. Particularly in the areas that could benefit from the optimisation of manufacturing and logistics processes to reduce the impact they have on the environment.

Sign up for the free insideBIGDATAnewsletter.

Join us on Twitter:@InsideBigData1 https://twitter.com/InsideBigData1

Read the original post:
AI Solving Real-world Problems and AI Ethics Among Top Trends for 2021, According to Oxylabs' AI and ML Advisory Board - insideBIGDATA