Integrative multi-omics identifies high risk multiple myeloma subgroup associated with significant DNA loss and dysregulated DNA repair and cell cycle pathways.

BACKGROUND: Despite significant therapeutic advances in improving lives of multiple myeloma (MM) patients, it remains mostly incurable, with patients ultimately becoming refractory to therapies. MM is a genetically heterogeneous disease and therapeutic resistance is driven by a complex interplay of disease pathobiology and mechanisms of drug resistance. We applied a multi-omics strategy using tumor-derived gene expression, single nucleotide variant, copy number variant, and structural variant profiles to investigate molecular subgroups in 514 newly diagnosed MM (NDMM) samples and identified 12 molecularly defined MM subgroups (MDMS1-12) with distinct genomic and transcriptomic features. RESULTS: Our integrative approach let us identify NDMM subgroups with transversal profiles to previously described ones, based on single data types, which shows the impact of this approach for disease stratification. One key novel subgroup is our MDMS8, associated with poor clinical outcome [median overall survival, 38 months (global log-rank p-value < 1 × 10-6)], which uniquely presents a broad genomic loss (> 9% of entire genome, t-test p value < 1e-5) driving dysregulation of various transcriptional programs affecting DNA repair and cell cycle/mitotic processes. This subgroup was validated on multiple independent datasets, and a master regulator analyses identified transcription factors controlling MDMS8 transcriptomic profile, including CKS1B and PRKDC among others, which are regulators of the DNA repair and cell cycle pathways. CONCLUSION: Using multi-omics unsupervised clustering we were able to discover a new high-risk multiple myeloma patient segment. This high-risk group presents diverse previously known genetic markers, but also a new characteristic defined by accumulation of genomic loss which seems to drive transcriptional dysregulation of cell cycle, DNA repair and DNA damage. Finally, our work identified various master regulators, including E2F2 and CKS1B as the genes controlling these key biological pathways.

Genetic and pharmacological interrogation of cancer vulnerability using a multiplexed cell line screening platform.

The multiplexed cancer cell line screening platform PRISM demonstrated its utility in testing hundreds of cell lines in a single run, possessing the potential to speed up anti-cancer drug discovery, validation and optimization. Here we described the development and implementation of a next-generation PRISM platform combining Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)/Cas9-mediated gene editing, cell line DNA barcoding and next-generation sequencing to enable genetic and/or pharmacological assessment of target addiction in hundreds of cell lines simultaneously. Both compound and CRISPR-knockout PRISM screens well recapitulated the results from individual assays and showed high consistency with a public database.

Deciphering the mechanisms of CC-122 resistance in DLBCL via a genome-wide CRISPR screen.

CC-122 is a next-generation cereblon E3 ligase-modulating agent that has demonstrated promising clinical efficacy in patients with relapsed or refractory diffuse large B-cell lymphoma (R/R DLBCL). Mechanistically, CC-122 induces the degradation of IKZF1/3, leading to T-cell activation and robust cell-autonomous killing in DLBCL. We report a genome-wide CRISPR/Cas9 screening for CC-122 in a DLBCL cell line SU-DHL-4 with follow-up mechanistic characterization in 6 DLBCL cell lines to identify genes regulating the response to CC-122. Top-ranked CC-122 resistance genes encode, not only well-defined members or regulators of the CUL4/DDB1/RBX1/CRBN E3 ubiquitin ligase complex, but also key components of signaling and transcriptional networks that have not been shown to modulate the response to cereblon modulators. Ablation of CYLD, NFKBIA, TRAF2, or TRAF3 induces hyperactivation of the canonical and/or noncanonical NF-κB pathways and subsequently diminishes CC-122-induced apoptosis in 5 of 6 DLBCL cell lines. Depletion of KCTD5, the substrate adaptor of the CUL3/RBX1/KCTD5 ubiquitin ligase complex, promotes the stabilization of its cognate substrate, GNG5, resulting in CC-122 resistance in HT, SU-DHL-4, and WSU-DLCL2. Furthermore, knockout of AMBRA1 renders resistance to CC-122 in SU-DHL-4 and U-2932, whereas knockout of RFX7 leads to resistance specifically in SU-DHL-4. The ubiquitous and cell line-specific mechanisms of CC-122 resistance in DLBCL cell lines revealed in this work pinpoint genetic alternations that are potentially associated with clinical resistance in patients and facilitate the development of biomarker strategies for patient stratification, which may improve clinical outcomes of patients with R/R DLBCL.

CC-90009, a novel cereblon E3 ligase modulator, targets acute myeloid leukemia blasts and leukemia stem cells.

A number of clinically validated drugs have been developed by repurposing the CUL4-DDB1-CRBN-RBX1 (CRL4CRBN) E3 ubiquitin ligase complex with molecular glue degraders to eliminate disease-driving proteins. Here, we present the identification of a first-in-class GSPT1-selective cereblon E3 ligase modulator, CC-90009. Biochemical, structural, and molecular characterization demonstrates that CC-90009 coopts the CRL4CRBN to selectively target GSPT1 for ubiquitination and proteasomal degradation. Depletion of GSPT1 by CC-90009 rapidly induces acute myeloid leukemia (AML) apoptosis, reducing leukemia engraftment and leukemia stem cells (LSCs) in large-scale primary patient xenografting of 35 independent AML samples, including those with adverse risk features. Using a genome-wide CRISPR-Cas9 screen for effectors of CC-90009 response, we uncovered the ILF2 and ILF3 heterodimeric complex as a novel regulator of cereblon expression. Knockout of ILF2/ILF3 decreases the production of full-length cereblon protein via modulating CRBN messenger RNA alternative splicing, leading to diminished response to CC-90009. The screen also revealed that the mTOR signaling and the integrated stress response specifically regulate the response to CC-90009 in contrast to other cereblon modulators. Hyperactivation of the mTOR pathway by inactivation of TSC1 and TSC2 protected against the growth inhibitory effect of CC-90009 by reducing CC-90009-induced binding of GSPT1 to cereblon and subsequent GSPT1 degradation. On the other hand, GSPT1 degradation promoted the activation of the GCN1/GCN2/ATF4 pathway and subsequent apoptosis in AML cells. Collectively, CC-90009 activity is mediated by multiple layers of signaling networks and pathways within AML blasts and LSCs, whose elucidation gives insight into further assessment of CC-90009s clinical utility. These trials were registered at www.clinicaltrials.gov as #NCT02848001 and #NCT04336982).

Community assessment to advance computational prediction of cancer drug combinations in a pharmacogenomic screen.

The effectiveness of most cancer targeted therapies is short-lived. Tumors often develop resistance that might be overcome with drug combinations. However, the number of possible combinations is vast, necessitating data-driven approaches to find optimal patient-specific treatments. Here we report AstraZeneca's large drug combination dataset, consisting of 11,576 experiments from 910 combinations across 85 molecularly characterized cancer cell lines, and results of a DREAM Challenge to evaluate computational strategies for predicting synergistic drug pairs and biomarkers. 160 teams participated to provide a comprehensive methodological development and benchmarking. Winning methods incorporate prior knowledge of drug-target interactions. Synergy is predicted with an accuracy matching biological replicates for >60% of combinations. However, 20% of drug combinations are poorly predicted by all methods. Genomic rationale for synergy predictions are identified, including ADAM17 inhibitor antagonism when combined with PIK3CB/D inhibition contrasting to synergy when combined with other PI3K-pathway inhibitors in PIK3CA mutant cells.

UBE2G1 governs the destruction of cereblon neomorphic substrates.

The cereblon modulating agents (CMs) including lenalidomide, pomalidomide and CC-220 repurpose the Cul4-RBX1-DDB1-CRBN (CRL4CRBN) E3 ubiquitin ligase complex to induce the degradation of specific neomorphic substrates via polyubiquitination in conjunction with E2 ubiquitin-conjugating enzymes, which have until now remained elusive. Here we show that the ubiquitin-conjugating enzymes UBE2G1 and UBE2D3 cooperatively promote the K48-linked polyubiquitination of CRL4CRBN neomorphic substrates via a sequential ubiquitination mechanism. Blockade of UBE2G1 diminishes the ubiquitination and degradation of neomorphic substrates, and consequent antitumor activities elicited by all tested CMs. For example, UBE2G1 inactivation significantly attenuated the degradation of myeloma survival factors IKZF1 and IKZF3 induced by lenalidomide and pomalidomide, hence conferring drug resistance. UBE2G1-deficient myeloma cells, however, remained sensitive to a more potent IKZF1/3 degrader CC-220. Collectively, it will be of fundamental interest to explore if loss of UBE2G1 activity is linked to clinical resistance to drugs that hijack the CRL4CRBN to eliminate disease-driving proteins.

Database of genomic biomarkers for cancer drugs and clinical targetability in solid tumors.

SUMMARY: Comprehensive genomic profiling is expected to revolutionize cancer therapy. In this Prospective, we present the prevalence of mutations and copy-number alterations with predictive associations across solid tumors at different levels of stringency for gene-drug targetability. More than 90% of The Cancer Genome Atlas samples have potentially targetable alterations, the majority with multiple events, illustrating the challenges for treatment prioritization given the complexity of the genomic landscape. Nearly 80% of the variants in rarely mutated oncogenes are of uncertain functional significance, reflecting the gap in our understanding of the relevance of many alterations potentially linked to therapeutic actions. Access to targeted agents in early clinical trials could affect treatment decision in 75% of patients with cancer. Prospective implementation of large-scale molecular profiling and standardized reports of predictive biomarkers are fundamental steps for making precision cancer medicine a reality.

Stepwise group sparse regression (SGSR): gene-set-based pharmacogenomic predictive models with stepwise selection of functional priors.

Complex mechanisms involving genomic aberrations in numerous proteins and pathways are believed to be a key cause of many diseases such as cancer. With recent advances in genomics, elucidating the molecular basis of cancer at a patient level is now feasible, and has led to personalized treatment strategies whereby a patient is treated according to his or her genomic profile. However, there is growing recognition that existing treatment modalities are overly simplistic, and do not fully account for the deep genomic complexity associated with sensitivity or resistance to cancer therapies. To overcome these limitations, large-scale pharmacogenomic screens of cancer cell lines--in conjunction with modern statistical learning approaches--have been used to explore the genetic underpinnings of drug response. While these analyses have demonstrated the ability to infer genetic predictors of compound sensitivity, to date most modeling approaches have been data-driven, i.e. they do not explicitly incorporate domain-specific knowledge (priors) in the process of learning a model. While a purely data-driven approach offers an unbiased perspective of the data--and may yield unexpected or novel insights--this strategy introduces challenges for both model interpretability and accuracy. In this study, we propose a novel prior-incorporated sparse regression model in which the choice of informative predictor sets is carried out by knowledge-driven priors (gene sets) in a stepwise fashion. Under regularization in a linear regression model, our algorithm is able to incorporate prior biological knowledge across the predictive variables thereby improving the interpretability of the final model with no loss--and often an improvement--in predictive performance. We evaluate the performance of our algorithm compared to well-known regularization methods such as LASSO, Ridge and Elastic net regression in the Cancer Cell Line Encyclopedia (CCLE) and Genomics of Drug Sensitivity in Cancer (Sanger) pharmacogenomics datasets, demonstrating that incorporation of the biological priors selected by our model confers improved predictability and interpretability, despite much fewer predictors, over existing state-of-the-art methods.

Functional kinomics identifies candidate therapeutic targets in head and neck cancer.

PURPOSE: To identify novel therapeutic drug targets for p53-mutant head and neck squamous cell carcinoma (HNSCC). EXPERIMENTAL DESIGN: RNAi kinome viability screens were performed on HNSCC cells, including autologous pairs from primary tumor and recurrent/metastatic lesions, and in parallel on murine squamous cell carcinoma (MSCC) cells derived from tumors of inbred mice bearing germline mutations in Trp53, and p53 regulatory genes: Atm, Prkdc, and p19(Arf). Cross-species analysis of cell lines stratified by p53 mutational status and metastatic phenotype was used to select 38 kinase targets. Both primary and secondary RNAi validation assays were performed on additional HNSCC cell lines to credential these kinase targets using multiple phenotypic endpoints. Kinase targets were also examined via chemical inhibition using a panel of kinase inhibitors. A preclinical study was conducted on the WEE1 kinase inhibitor, MK-1775. RESULTS: Our functional kinomics approach identified novel survival kinases in HNSCC involved in G2-M cell-cycle checkpoint, SFK, PI3K, and FAK pathways. RNAi-mediated knockdown and chemical inhibition of the WEE1 kinase with a specific inhibitor, MK-1775, had a significant effect on both viability and apoptosis. Sensitivity to the MK-1775 kinase inhibitor is in part determined by p53 mutational status, and due to unscheduled mitotic entry. MK-1775 displays single-agent activity and potentiates the efficacy of cisplatin in a p53-mutant HNSCC xenograft model. CONCLUSIONS: WEE1 kinase is a potential therapeutic drug target for HNSCC. This study supports the application of a functional kinomics strategy to identify novel therapeutic targets for cancer.

Synthetic lethal screens as a means to understand and treat MYC-driven cancers.

Although therapeutics against MYC could potentially be used against a wide range of human cancers, MYC-targeted therapies have proven difficult to develop. The convergence of breakthroughs in human genomics and in gene silencing using RNA interference (RNAi) have recently allowed functional interrogation of the genome and systematic identification of synthetic lethal interactions with hyperactive MYC. Here, we focus on the pathways that have emerged through RNAi screens and present evidence that a subset of genes showing synthetic lethality with MYC are significantly interconnected and linked to chromatin and transcriptional processes, as well as to DNA repair and cell cycle checkpoints. Other synthetic lethal interactions with MYC point to novel pathways and potentially broaden the repertoire of targeted therapies. The elucidation of MYC synthetic lethal interactions is still in its infancy, and how these interactions may be influenced by tissue-specific programs and by concurrent genetic change will require further investigation. Nevertheless, we predict that these studies may lead the way to novel therapeutic approaches and new insights into the role of MYC in cancer.

The Stream algorithm: computationally efficient ridge-regression via Bayesian model averaging, and applications to pharmacogenomic prediction of cancer cell line sensitivity.

Computational efficiency is important for learning algorithms operating in the "large p, small n" setting. In computational biology, the analysis of data sets containing tens of thousands of features ("large p"), but only a few hundred samples ("small n"), is nowadays routine, and regularized regression approaches such as ridge-regression, lasso, and elastic-net are popular choices. In this paper we propose a novel and highly efficient Bayesian inference method for fitting ridge-regression. Our method is fully analytical, and bypasses the need for expensive tuning parameter optimization, via cross-validation, by employing Bayesian model averaging over the grid of tuning parameters. Additional computational efficiency is achieved by adopting the singular value decomposition reparametrization of the ridge-regression model, replacing computationally expensive inversions of large p × p matrices by efficient inversions of small and diagonal n × n matrices. We show in simulation studies and in the analysis of two large cancer cell line data panels that our algorithm achieves slightly better predictive performance than cross-validated ridge-regression while requiring only a fraction of the computation time. Furthermore, in comparisons based on the cell line data sets, our algorithm systematically out-performs the lasso in both predictive performance and computation time, and shows equivalent predictive performance, but considerably smaller computation time, than the elastic-net.

Systematic assessment of analytical methods for drug sensitivity prediction from cancer cell line data.

Large-scale pharmacogenomic screens of cancer cell lines have emerged as an attractive pre-clinical system for identifying tumor genetic subtypes with selective sensitivity to targeted therapeutic strategies. Application of modern machine learning approaches to pharmacogenomic datasets have demonstrated the ability to infer genomic predictors of compound sensitivity. Such modeling approaches entail many analytical design choices; however, a systematic study evaluating the relative performance attributable to each design choice is not yet available. In this work, we evaluated over 110,000 different models, based on a multifactorial experimental design testing systematic combinations of modeling factors within several categories of modeling choices, including: type of algorithm, type of molecular feature data, compound being predicted, method of summarizing compound sensitivity values, and whether predictions are based on discretized or continuous response values. Our results suggest that model input data (type of molecular features and choice of compound) are the primary factors explaining model performance, followed by choice of algorithm. Our results also provide a statistically principled set of recommended modeling guidelines, including: using elastic net or ridge regression with input features from all genomic profiling platforms, most importantly, gene expression features, to predict continuous-valued sensitivity scores summarized using the area under the dose response curve, with pathway targeted compounds most likely to yield the most accurate predictors. In addition, our study provides a publicly available resource of all modeling results, an open source code base, and experimental design for researchers throughout the community to build on our results and assess novel methodologies or applications in related predictive modeling problems.

Improving breast cancer survival analysis through competition-based multidimensional modeling.

Breast cancer is the most common malignancy in women and is responsible for hundreds of thousands of deaths annually. As with most cancers, it is a heterogeneous disease and different breast cancer subtypes are treated differently. Understanding the difference in prognosis for breast cancer based on its molecular and phenotypic features is one avenue for improving treatment by matching the proper treatment with molecular subtypes of the disease. In this work, we employed a competition-based approach to modeling breast cancer prognosis using large datasets containing genomic and clinical information and an online real-time leaderboard program used to speed feedback to the modeling team and to encourage each modeler to work towards achieving a higher ranked submission. We find that machine learning methods combined with molecular features selected based on expert prior knowledge can improve survival predictions compared to current best-in-class methodologies and that ensemble models trained across multiple user submissions systematically outperform individual models within the ensemble. We also find that model scores are highly consistent across multiple independent evaluations. This study serves as the pilot phase of a much larger competition open to the whole research community, with the goal of understanding general strategies for model optimization using clinical and molecular profiling data and providing an objective, transparent system for assessing prognostic models.

hARACNe: improving the accuracy of regulatory model reverse engineering via higher-order data processing inequality tests.

A key goal of systems biology is to elucidate molecular mechanisms associated with physiologic and pathologic phenotypes based on the systematic and genome-wide understanding of cell context-specific molecular interaction models. To this end, reverse engineering approaches have been used to systematically dissect regulatory interactions in a specific tissue, based on the availability of large molecular profile datasets, thus improving our mechanistic understanding of complex diseases, such as cancer. In this paper, we introduce high-order Algorithm for the Reconstruction of Accurate Cellular Network (hARACNe), an extension of the ARACNe algorithm for the dissection of transcriptional regulatory networks. ARACNe uses the data processing inequality (DPI), from information theory, to detect and prune indirect interactions that are unlikely to be mediated by an actual physical interaction. Whereas ARACNe considers only first-order indirect interactions, i.e. those mediated by only one extra regulator, hARACNe considers a generalized form of indirect interactions via two, three or more other regulators. We show that use of higher-order DPI resulted in significantly improved performance, based on transcription factor (TF)-specific ChIP-chip data, as well as on gene expression profile following RNAi-mediated TF silencing.