Loss of NEDD8 in cancer cells causes vulnerability to immune checkpoint blockade in triple-negative breast cancer.

Immune checkpoint blockade therapy aims to activate the immune system to eliminate cancer cells. However, clinical benefits are only recorded in a subset of patients. Here, we leverage genome-wide CRISPR/Cas9 screens in a Tumor-Immune co-Culture System focusing on triple-negative breast cancer (TNBC). We reveal that NEDD8 loss in cancer cells causes a vulnerability to nivolumab (anti-PD-1). Genetic deletion of NEDD8 only delays cell division initially but cell proliferation is unaffected after recovery. Since the NEDD8 gene is commonly essential, we validate this observation with additional CRISPR screens and uncover enhanced immunogenicity in NEDD8 deficient cells using proteomics. In female immunocompetent mice, PD-1 blockade lacks efficacy against established EO771 breast cancer tumors. In contrast, we observe tumor regression mediated by CD8+ T cells against Nedd8 deficient EO771 tumors after PD-1 blockade. In essence, we provide evidence that NEDD8 is conditionally essential in TNBC and presents as a synergistic drug target for PD-1/L1 blockade therapy.

The Oncology Biomarker Discovery framework reveals cetuximab and bevacizumab response patterns in metastatic colorectal cancer.

Precision medicine has revolutionised cancer treatments; however, actionable biomarkers remain scarce. To address this, we develop the Oncology Biomarker Discovery (OncoBird) framework for analysing the molecular and biomarker landscape of randomised controlled clinical trials. OncoBird identifies biomarkers based on single genes or mutually exclusive genetic alterations in isolation or in the context of tumour subtypes, and finally, assesses predictive components by their treatment interactions. Here, we utilise the open-label, randomised phase III trial (FIRE-3, AIO KRK-0306) in metastatic colorectal carcinoma patients, who received either cetuximab or bevacizumab in combination with 5-fluorouracil, folinic acid and irinotecan (FOLFIRI). We systematically identify five biomarkers with predictive components, e.g., patients with tumours that carry chr20q amplifications or lack mutually exclusive ERK signalling mutations benefited from cetuximab compared to bevacizumab. In summary, OncoBird characterises the molecular landscape and outlines actionable biomarkers, which generalises to any molecularly characterised randomised controlled trial.

Oncogene-induced MALT1 protease activity drives posttranscriptional gene expression in malignant lymphomas.

Constitutive mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) activity drives survival of malignant lymphomas addicted to chronic B-cell receptor signaling, oncogenic CARD11, or the API2-MALT1 (also BIRC3::MALT1) fusion oncoprotein. Although MALT1 scaffolding induces NF-κB-dependent survival signaling, MALT1 protease function is thought to augment NF-κB activation by cleaving signaling mediators and transcriptional regulators in B-cell lymphomas. However, the pathological role of MALT1 protease function in lymphomagenesis is not well understood. Here, we show that TRAF6 controls MALT1-dependent activation of NF-κB transcriptional responses but is dispensable for MALT1 protease activation driven by oncogenic CARD11. To uncouple enzymatic and nonenzymatic functions of MALT1, we analyzed TRAF6-dependent and -independent as well as MALT1 protease-dependent gene expression profiles downstream of oncogenic CARD11 and API2-MALT1. The data suggest that by cleaving and inactivating the RNA binding proteins Regnase-1 and Roquin-1/2, MALT1 protease induces posttranscriptional upregulation of many genes including NFKBIZ/IκBζ, NFKBID/IκBNS, and ZC3H12A/Regnase-1 in activated B-cell-like diffuse large B-cell lymphoma (ABC DLBCL). We demonstrate that oncogene-driven MALT1 activity in ABC DLBCL cells regulates NFKBIZ and NFKBID induction on an mRNA level via releasing a brake imposed by Regnase-1 and Roquin-1/2. Furthermore, MALT1 protease drives posttranscriptional gene induction in the context of the API2-MALT1 fusion created by the recurrent t(11;18)(q21;q21) translocation in MALT lymphoma. Thus, MALT1 paracaspase acts as a bifurcation point for enhancing transcriptional and posttranscriptional gene expression in malignant lymphomas. Moreover, the identification of MALT1 protease-selective target genes provides specific biomarkers for the clinical evaluation of MALT1 inhibitors.

Inferred Ancestral Origin of Cancer Cell Lines Associates with Differential Drug Response.

Disparities between risk, treatment outcomes and survival rates in cancer patients across the world may be attributed to socioeconomic factors. In addition, the role of ancestry is frequently discussed. In preclinical studies, high-throughput drug screens in cancer cell lines have empowered the identification of clinically relevant molecular biomarkers of drug sensitivity; however, the genetic ancestry from tissue donors has been largely neglected in this setting. In order to address this, here, we show that the inferred ancestry of cancer cell lines is conserved and may impact drug response in patients as a predictive covariate in high-throughput drug screens. We found that there are differential drug responses between European and East Asian ancestries, especially when treated with PI3K/mTOR inhibitors. Our finding emphasizes a new angle in precision medicine, as cancer intervention strategies should consider the germline landscape, thereby reducing the failure rate of clinical trials.

A statistical framework for assessing pharmacological responses and biomarkers using uncertainty estimates.

High-throughput testing of drugs across molecular-characterised cell lines can identify candidate treatments and discover biomarkers. However, the cells' response to a drug is typically quantified by a summary statistic from a best-fit dose-response curve, whilst neglecting the uncertainty of the curve fit and the potential variability in the raw readouts. Here, we model the experimental variance using Gaussian Processes, and subsequently, leverage uncertainty estimates to identify associated biomarkers with a new Bayesian framework. Applied to in vitro screening data on 265 compounds across 1074 cancer cell lines, our models identified 24 clinically established drug-response biomarkers, and provided evidence for six novel biomarkers by accounting for association with low uncertainty. We validated our uncertainty estimates with an additional drug screen of 26 drugs, 10 cell lines with 8 to 9 replicates. Our method is applicable to any dose-response data without replicates, and improves biomarker discovery for precision medicine.

Identification of Intrinsic Drug Resistance and Its Biomarkers in High-Throughput Pharmacogenomic and CRISPR Screens.

High-throughput drug screens in cancer cell lines test compounds at low concentrations, thereby enabling the identification of drug-sensitivity biomarkers, while resistance biomarkers remain underexplored. Dissecting meaningful drug responses at high concentrations is challenging due to cytotoxicity, i.e., off-target effects, thus limiting resistance biomarker discovery to frequently mutated cancer genes. To address this, we interrogate subpopulations carrying sensitivity biomarkers and consecutively investigate unexpectedly resistant (UNRES) cell lines for unique genetic alterations that may drive resistance. By analyzing the GDSC and CTRP datasets, we find 53 and 35 UNRES cases, respectively. For 24 and 28 of them, we highlight putative resistance biomarkers. We find clinically relevant cases such as EGFRT790M mutation in NCI-H1975 or PTEN loss in NCI-H1650 cells, in lung adenocarcinoma treated with EGFR inhibitors. Interrogating the underpinnings of drug resistance with publicly available CRISPR phenotypic assays assists in prioritizing resistance drivers, offering hypotheses for drug combinations.

Concise Polygenic Models for Cancer-Specific Identification of Drug-Sensitive Tumors from Their Multi-Omics Profiles.

In silico models to predict which tumors will respond to a given drug are necessary for Precision Oncology. However, predictive models are only available for a handful of cases (each case being a given drug acting on tumors of a specific cancer type). A way to generate predictive models for the remaining cases is with suitable machine learning algorithms that are yet to be applied to existing in vitro pharmacogenomics datasets. Here, we apply XGBoost integrated with a stringent feature selection approach, which is an algorithm that is advantageous for these high-dimensional problems. Thus, we identified and validated 118 predictive models for 62 drugs across five cancer types by exploiting four molecular profiles (sequence mutations, copy-number alterations, gene expression, and DNA methylation). Predictive models were found in each cancer type and with every molecular profile. On average, no omics profile or cancer type obtained models with higher predictive accuracy than the rest. However, within a given cancer type, some molecular profiles were overrepresented among predictive models. For instance, CNA profiles were predictive in breast invasive carcinoma (BRCA) cell lines, but not in small cell lung cancer (SCLC) cell lines where gene expression (GEX) and DNA methylation profiles were the most predictive. Lastly, we identified the best XGBoost model per cancer type and analyzed their selected features. For each model, some of the genes in the selected list had already been found to be individually linked to the response to that drug, providing additional evidence of the usefulness of these models and the merits of the feature selection scheme.

Stratification and prediction of drug synergy based on target functional similarity.

Drug combinations can expand therapeutic options and address cancer's resistance. However, the combinatorial space is enormous precluding its systematic exploration. Therefore, synergy prediction strategies are essential. We here present an approach to prioritise drug combinations in high-throughput screens and to stratify synergistic responses. At the core of our approach is the observation that the likelihood of synergy increases when targeting proteins with either strong functional similarity or dissimilarity. We estimate the similarity applying a multitask machine learning approach to basal gene expression and response to single drugs. We tested 7 protein target pairs (representing 29 combinations) and predicted their synergies in 33 breast cancer cell lines. In addition, we experimentally validated predicted synergy of the BRAF/insulin receptor combination (Dabrafenib/BMS-754807) in 48 colorectal cancer cell lines. We anticipate that our approaches can be used for prioritization of drug combinations in large scale screenings, and to maximize the efficacy of drugs already known to induce synergy, ultimately enabling patient stratification.

Quantitative Proteome Landscape of the NCI-60 Cancer Cell Lines.

Here we describe a proteomic data resource for the NCI-60 cell lines generated by pressure cycling technology and SWATH mass spectrometry. We developed the DIA-expert software to curate and visualize the SWATH data, leading to reproducible detection of over 3,100 SwissProt proteotypic proteins and systematic quantification of pathway activities. Stoichiometric relationships of interacting proteins for DNA replication, repair, the chromatin remodeling NuRD complex, ß-catenin, RNA metabolism, and prefoldins are more evident than that at the mRNA level. The data are available in CellMiner (discover.nci.nih.gov/cellminercdb and discover.nci.nih.gov/cellminer), allowing casual users to test hypotheses and perform integrative, cross-database analyses of multi-omic drug response correlations for over 20,000 drugs. We demonstrate the value of proteome data in predicting drug response for over 240 clinically relevant chemotherapeutic and targeted therapies. In summary, we present a novel proteome resource for the NCI-60, together with relevant software tools, and demonstrate the benefit of proteome analyses.

Defining subpopulations of differential drug response to reveal novel target populations.

Personalised medicine has predominantly focused on genetically altered cancer genes that stratify drug responses, but there is a need to objectively evaluate differential pharmacology patterns at a subpopulation level. Here, we introduce an approach based on unsupervised machine learning to compare the pharmacological response relationships between 327 pairs of cancer therapies. This approach integrated multiple measures of response to identify subpopulations that react differently to inhibitors of the same or different targets to understand mechanisms of resistance and pathway cross-talk. MEK, BRAF, and PI3K inhibitors were shown to be effective as combination therapies for particular BRAF mutant subpopulations. A systematic analysis of preclinical data for a failed phase III trial of selumetinib combined with docetaxel in lung cancer suggests potential indications in pancreatic and colorectal cancers with KRAS mutation. This data-informed study exemplifies a method for stratified medicine to identify novel cancer subpopulations, their genetic biomarkers, and effective drug combinations.

Applying synergy metrics to combination screening data: agreements, disagreements and pitfalls.

Synergistic drug combinations are commonly sought to overcome monotherapy resistance in cancer treatment. To identify such combinations, high-throughput cancer cell line combination screens are performed; and synergy is quantified using competing models based on fundamentally different assumptions. Here, we compare the behaviour of four synergy models, namely Loewe additivity, Bliss independence, highest single agent and zero interaction potency, using the Merck oncology combination screen. We evaluate agreements and disagreements between the models and investigate putative artefacts of each model's assumptions. Despite at least moderate concordance between scores (Pearson's r >0.32, Spearman's ρ>0.34), multiple instances of strong disagreement were observed. Those disagreements are driven by, among others, large differences in tested concentrations, maximum response values and median effective concentrations.

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.

The germline genetic component of drug sensitivity in cancer cell lines.

Patients with seemingly the same tumour can respond very differently to treatment. There are strong, well-established effects of somatic mutations on drug efficacy, but there is at-most anecdotal evidence of a germline component to drug response. Here, we report a systematic survey of how inherited germline variants affect drug susceptibility in cancer cell lines. We develop a joint analysis approach that leverages both germline and somatic variants, before applying it to screening data from 993 cell lines and 265 drugs. Surprisingly, we find that the germline contribution to variation in drug susceptibility can be as large or larger than effects due to somatic mutations. Several of the associations identified have a direct relationship to the drug target. Finally, using 17-AAG response as an example, we show how germline effects in combination with transcriptomic data can be leveraged for improved patient stratification and to identify new markers for drug sensitivity.

GDSCTools for mining pharmacogenomic interactions in cancer.

Motivation: Large pharmacogenomic screenings integrate heterogeneous cancer genomic datasets as well as anti-cancer drug responses on thousand human cancer cell lines. Mining this data to identify new therapies for cancer sub-populations would benefit from common data structures, modular computational biology tools and user-friendly interfaces. Results: We have developed GDSCTools: a software aimed at the identification of clinically relevant genomic markers of drug response. The Genomics of Drug Sensitivity in Cancer (GDSC) database (www.cancerRxgene.org) integrates heterogeneous cancer genomic datasets as well as anti-cancer drug responses on a thousand cancer cell lines. Including statistical tools (analysis of variance) and predictive methods (Elastic Net), as well as common data structures, GDSCTools allows users to reproduce published results from GDSC and to implement new analytical methods. In addition, non-GDSC data resources can also be analysed since drug responses and genomic features can be encoded as CSV files. Contact: thomas.cokelaer@pasteur.fr or saezrodriguez.rwth-aachen.de or mg12@sanger.ac.uk. Supplementary information: Supplementary data are available at Bioinformatics online.

A Landscape of Pharmacogenomic Interactions in Cancer.

Systematic studies of cancer genomes have provided unprecedented insights into the molecular nature of cancer. Using this information to guide the development and application of therapies in the clinic is challenging. Here, we report how cancer-driven alterations identified in 11,289 tumors from 29 tissues (integrating somatic mutations, copy number alterations, DNA methylation, and gene expression) can be mapped onto 1,001 molecularly annotated human cancer cell lines and correlated with sensitivity to 265 drugs. We find that cell lines faithfully recapitulate oncogenic alterations identified in tumors, find that many of these associate with drug sensitivity/resistance, and highlight the importance of tissue lineage in mediating drug response. Logic-based modeling uncovers combinations of alterations that sensitize to drugs, while machine learning demonstrates the relative importance of different data types in predicting drug response. Our analysis and datasets are rich resources to link genotypes with cellular phenotypes and to identify therapeutic options for selected cancer sub-populations.

A community effort to assess and improve drug sensitivity prediction algorithms.

Predicting the best treatment strategy from genomic information is a core goal of precision medicine. Here we focus on predicting drug response based on a cohort of genomic, epigenomic and proteomic profiling data sets measured in human breast cancer cell lines. Through a collaborative effort between the National Cancer Institute (NCI) and the Dialogue on Reverse Engineering Assessment and Methods (DREAM) project, we analyzed a total of 44 drug sensitivity prediction algorithms. The top-performing approaches modeled nonlinear relationships and incorporated biological pathway information. We found that gene expression microarrays consistently provided the best predictive power of the individual profiling data sets; however, performance was increased by including multiple, independent data sets. We discuss the innovations underlying the top-performing methodology, Bayesian multitask MKL, and we provide detailed descriptions of all methods. This study establishes benchmarks for drug sensitivity prediction and identifies approaches that can be leveraged for the development of new methods.

Machine learning prediction of cancer cell sensitivity to drugs based on genomic and chemical properties.

Predicting the response of a specific cancer to a therapy is a major goal in modern oncology that should ultimately lead to a personalised treatment. High-throughput screenings of potentially active compounds against a panel of genomically heterogeneous cancer cell lines have unveiled multiple relationships between genomic alterations and drug responses. Various computational approaches have been proposed to predict sensitivity based on genomic features, while others have used the chemical properties of the drugs to ascertain their effect. In an effort to integrate these complementary approaches, we developed machine learning models to predict the response of cancer cell lines to drug treatment, quantified through IC50 values, based on both the genomic features of the cell lines and the chemical properties of the considered drugs. Models predicted IC50 values in a 8-fold cross-validation and an independent blind test with coefficient of determination R² of 0.72 and 0.64 respectively. Furthermore, models were able to predict with comparable accuracy (R² of 0.61) IC50s of cell lines from a tissue not used in the training stage. Our in silico models can be used to optimise the experimental design of drug-cell screenings by estimating a large proportion of missing IC50 values rather than experimentally measuring them. The implications of our results go beyond virtual drug screening design: potentially thousands of drugs could be probed in silico to systematically test their potential efficacy as anti-tumour agents based on their structure, thus providing a computational framework to identify new drug repositioning opportunities as well as ultimately be useful for personalized medicine by linking the genomic traits of patients to drug sensitivity.

Adaptive informatics for multifactorial and high-content biological data.

Whereas genomic data are universally machine-readable, data from imaging, multiplex biochemistry, flow cytometry and other cell- and tissue-based assays usually reside in loosely organized files of poorly documented provenance. This arises because the relational databases used in genomic research are difficult to adapt to rapidly evolving experimental designs, data formats and analytic algorithms. Here we describe an adaptive approach to managing experimental data based on semantically typed data hypercubes (SDCubes) that combine hierarchical data format 5 (HDF5) and extensible markup language (XML) file types. We demonstrate the application of SDCube-based storage using ImageRail, a software package for high-throughput microscopy. Experimental design and its day-to-day evolution, not rigid standards, determine how ImageRail data are organized in SDCubes. We applied ImageRail to collect and analyze drug dose-response landscapes in human cell lines at single-cell resolution.