Harnessing Connectivity in a Large-Scale Small-Molecule Sensitivity Dataset

Deep learning for drug response prediction in cancer

Predicting the sensitivity of tumors to specific anti-cancer treatments is a challenge of paramount importance for precision medicine. Machine learning(ML) algorithms can be trained on high-throughput screening data to develop models that are able to predict the response of cancer cell lines and patients to novel drugs or drug combinations. Deep learning (DL) refers to a distinct class of ML algorithms that have achieved top-level performance in a variety of fields, including drug discovery. These types of models have unique characteristics that may make them more suitable for the complex task of modeling drug response based on both biological and chemical data, but the application of DL to drug response prediction has been unexplored until very recently. The few studies that have been published have shown promising results, and the use of DL for drug response prediction is beginning to attract greater interest from researchers in the field. In this article, we critically review recently published studies that have employed DL methods to predict drug response in cancer cell lines. We also provide a brief description of DL and the main types of architectures that have been used in these studies. Additionally, we present a selection of publicly available drug screening data resources that can be used to develop drug response prediction models. Finally, we also address the limitations of these approaches and provide a discussion on possible paths for further improvement. Contact:  mrocha@di.uminho.pt.

Epigenetic Silencing of CDR1as Drives IGF2BP3-Mediated Melanoma Invasion and Metastasis

Metastasis is the primary cause of death of cancer patients. Dissecting mechanisms governing metastatic spread may uncover important tumor biology and/or yield promising therapeutic insights. Here, we investigated the role of circular RNAs (circRNA) in metastasis, using melanoma as a model aggressive tumor. We identified silencing of cerebellar degeneration-related 1 antisense (CDR1as), a regulator of miR-7, as a hallmark of melanoma progression. CDR1as depletion results from epigenetic silencing of LINC00632, its originating long non-coding RNA (lncRNA) and promotes invasion in vitro and metastasis in vivo through a miR-7-independent, IGF2BP3-mediated mechanism. Moreover, CDR1as levels reflect cellular states associated with distinct therapeutic responses. Our study reveals functional, prognostic, and predictive roles for CDR1as and expose circRNAs as key players in metastasis.

Do We have a Satisfactory Cell Viability Assay? Review of the Currently Commercially-Available Assays

Cell-based assays are an important part of the drug discovery process and clinical research. One of the main hurdles is to design sufficiently robust assays with adequate signal to noise parameters while maintaining the inherent physiology of the cells and not interfering with the pharmacology of target being investigated. A plethora of assays that assess cell viability (or cell heath in general) are commercially available and can be classified under different categories according to their concepts and principle of reactions. The assays are valuable tools, however, suffer from a large number of limitations. Some of these limitations can be procedural or operational, but others can be critical as those related to a poor concept or the lack of proof of concept of an assay, e.g. those relying on differential permeability of dyes in-and-out of viable versus compromised cell membranes. While the assays can differentiate between dead and live cells, most, if not all, of them can just assess the relative performance of cells rather than providing a clear distinction between healthy and dying cells. The possible impact of relatively high molecular weight dyes, used in most of the assay, on cell viability has not been addressed. More innovative assays are needed, and until better alternatives are developed, setup of current cell-based studies and data interpretation should be made with the limitations in mind. Negative and positive control should be considered whenever feasible. Also, researchers should use more than one orthogonal method for better assessment of cell health.

Histone 2A Family Member J Drives Mesenchymal Transition and Temozolomide Resistance in Glioblastoma Multiforme

Glioblastoma multiforme (GBM) is the most aggressive brain tumor and has a poor prognosis and is poorly sensitive to radiotherapy or temozolomide (TMZ) chemotherapy. Therefore, identifying new biomarkers to predict therapeutic responses of GBM is urgently needed. By using The Cancer Genome Atlas (TCGA) database, we found that the upregulation of histone 2A family member J (H2AFJ), but not other H2AFs, is extensively detected in the therapeutic-insensitive mesenchymal, IDH wildtype, MGMT unmethylated, or non-G-CIMP GBM and is associated with poor TMZ responsiveness independent of radiation. Similar views were also found in GBM cell lines. Whereas H2AFJ knockdown diminished TMZ resistance, H2AFJ overexpression promoted TMZ resistance in a panel of GBM cell lines. Gene set enrichment analysis (GSEA) revealed that H2AFJ upregulation accompanied by the activation of TNF-α/NF-κB and IL-6/STAT3-related pathways is highly predicted. Luciferase-based promoter activity assay further validated that the activities of NF-κB and STAT3 are causally affected by H2AFJ expression in GBM cells. Moreover, we found that therapeutic targeting HADC3 by tacedinaline or NF-κB by ML029 is likely able to overcome the TMZ resistance in GBM cells with H2AFJ upregulation. Significantly, the GBM cohorts harboring a high-level H2AFJ transcript combined with high-level expression of TNF-α/NF-κB geneset, IL-6/STAT3 geneset or HADC3 were associated with a shorter time to tumor repopulation after initial treatment with TMZ. These findings not only provide H2AFJ as a biomarker to predict TMZ therapeutic effectiveness but also suggest a new strategy to combat TMZ-insensitive GBM by targeting the interaction network constructed by TNF-α/NF-κB, IL-6/STAT3, HDAC3, and H2AFJ.

Improving drug response prediction by integrating multiple data sources: matrix factorization, kernel and network-based approaches

Predicting the response of cancer cell lines to specific drugs is one of the central problems in personalized medicine, where the cell lines show diverse characteristics. Researchers have developed a variety of computational methods to discover associations between drugs and cell lines, and improved drug sensitivity analyses by integrating heterogeneous biological data. However, choosing informative data sources and methods that can incorporate multiple sources efficiently is the challenging part of successful analysis in personalized medicine. The reason is that finding decisive factors of cancer and developing methods that can overcome the problems of integrating data, such as differences in data structures and data complexities, are difficult. In this review, we summarize recent advances in data integration-based machine learning for drug response prediction, by categorizing methods as matrix factorization-based, kernel-based and network-based methods. We also present a short description of relevant databases used as a benchmark in drug response prediction analyses, followed by providing a brief discussion of challenges faced in integrating and interpreting data from multiple sources. Finally, we address the advantages of combining multiple heterogeneous data sources on drug sensitivity analysis by showing an experimental comparison. Contact:  betul.guvenc@aalto.fi

Human Cell-Based in vitro Phenotypic Profiling for Drug Safety-Related Attrition

Ensuring the safety of new drugs is critically important to regulators, pharmaceutical researchers and patients alike. Even so, unexpected toxicities still account for 20–30% of clinical trial failures, in part due to the persistence of animal testing as the primary approach for de-risking new drugs. Clearly, improved methods for safety attrition that incorporate human-relevant biology are needed. This recognition has spurred interest in non-animal alternatives or new approach methodologies (NAMs) including in vitro models that utilize advances in the culture of human cell types to provide greater clinical relevance for assessing risk. These phenotypic assay systems use human primary and induced pluripotent stem cell-derived cells in various formats, including co-cultures and advanced cellular systems such as organoids, bioprinted tissues, and organs-on-a-chip. Despite the promise of these human-based phenotypic approaches, adoption of these platforms into drug discovery programs for reducing safety-related attrition has been slow. Here we discuss the value of large-scale human cell-based phenotypic profiling for incorporating human-specific biology into the de-risking process. We describe learnings from our experiences with human primary cell-based assays and analysis of clinically relevant reference datasets in developing in vitro-based toxicity signatures. We also describe how Adverse Outcome Pathway (AOP) frameworks can be used to integrate results from diverse platforms congruent with weight-of-evidence approaches from risk assessment to improve safety-related decisions in early discovery.

Genetic Interactions and Tissue Specificity Modulate the Association of Mutations with Drug Response

In oncology, biomarkers are widely used to predict subgroups of patients that respond to a given drug. Although clinical decisions often rely on single gene biomarkers, machine learning approaches tend to generate complex multi-gene biomarkers that are hard to interpret. Models predicting drug response based on multiple altered genes often assume that the effects of single alterations are independent. We asked whether the association of cancer driver mutations with drug response is modulated by other driver mutations or the tissue of origin. We developed an analytic framework based on linear regression to study interactions in pharmacogenomic data from two large cancer cell line panels. Starting from a model with only covariates, we included additional variables only if they significantly improved simpler models. This allows to systematically assess interactions in small, easily interpretable models. Our results show that including mutation-mutation interactions in drug response prediction models tends to improve model performance and robustness. For example, we found that TP53 mutations decrease sensitivity to BRAF inhibitors in BRAF-mutated cell lines and patient tumors, suggesting a therapeutic benefit of combining inhibition of oncogenic BRAF with reactivation of the tumor suppressor TP53. Moreover, we identified tissue-specific mutation-drug associations and synthetic lethal triplets where the simultaneous mutation of two genes sensitizes cells to a drug. In summary, our interaction-based approach contributes to a holistic view on the determining factors of drug response.

Signatures of cell death and proliferation in perturbation transcriptomics data-from confounding factor to effective prediction

Transcriptional perturbation signatures are valuable data sources for functional genomics. Linking perturbation signatures to screenings opens the possibility to model cellular phenotypes from expression data and to identify efficacious drugs. We linked perturbation transcriptomics data from the LINCS-L1000 project with cell viability information upon genetic (Achilles project) and chemical (CTRP screen) perturbations yielding more than 90 000 signature–viability pairs. An integrated analysis showed that the cell viability signature is a major factor underlying perturbation signatures. The signature is linked to transcription factors regulating cell death, proliferation and division time. We used the cell viability–signature relationship to predict viability from transcriptomics signatures, and identified and validated compounds that induce cell death in tumor cell lines. We showed that cellular toxicity can lead to unexpected similarity of signatures, confounding mechanism of action discovery. Consensus compound signatures predicted cell-specific drug sensitivity, even if the signature is not measured in the same cell line, and outperformed conventional drug-specific features. Our results can help in understanding mechanisms behind cell death and removing confounding factors of transcriptomic perturbation screens. To interactively browse our results and predict cell viability in new gene expression samples, we developed CEVIChE (CEll VIability Calculator from gene Expression; https://saezlab.shinyapps.io/ceviche/).

Clonal selection confers distinct evolutionary trajectories in BRAF-driven cancers

Molecular determinants governing the evolution of tumor subclones toward phylogenetic branches or fixation remain unknown. Using sequencing data, we model the propagation and selection of clones expressing distinct categories of BRAF mutations to estimate their evolutionary trajectories. We show that strongly activating BRAF mutations demonstrate hard sweep dynamics, whereas mutations with less pronounced activation of the BRAF signaling pathway confer soft sweeps or are subclonal. We use clonal reconstructions to estimate the strength of "driver" selection in individual tumors. Using tumors cells and human-derived murine xenografts, we show that tumor sweep dynamics can significantly affect responses to targeted inhibitors of BRAF/MEK or DNA damaging agents. Our study uncovers patterns of distinct BRAF clonal evolutionary dynamics and nominates therapeutic strategies based on the identity of the BRAF mutation and its clonal composition.

Chromatin regulators mediate anthracycline sensitivity in breast cancer

Anthracyclines are a highly effective component of curative breast cancer chemotherapy, but are associated with significant morbidity^1,2. Since anthracyclines work in part via inhibition of topoisomerase-II (TOP2) on accessible DNA^3,4, we hypothesized that chromatin regulatory genes (CRGs) that mediate DNA accessibility might predict anthracycline response. We elucidate the role of CRGs in anthracycline sensitivity in breast cancer through integrative analysis of patient and cell line data. We identify a consensus set of 38 CRGs associated with anthracycline response across ten cell line datasets. Evaluating the interaction between expression and treatment in predicting survival in a metacohort of 1006 early-stage breast cancer patients, we identify 54 CRGs whose expression levels dictate anthracycline benefit across the clinical subgroups, 12 of which overlapped with those identified in vitro. CRGs that promote DNA accessibility, including Trithorax complex members, were associated with anthracycline sensitivity when highly expressed, whereas CRGs that reduce accessibility such as Polycomb complex proteins, were associated with decreased anthracycline sensitivity. We show that KDM4B modulates TOP2 accessibility to chromatin, elucidating a mechanism of TOP2 inhibitor sensitivity. These findings indicate that CRGs mediate anthracycline benefit by modulating DNA accessibility with implications for breast cancer patient stratification and treatment decision making.

Network-based Biased Tree Ensembles (NetBiTE) for Drug Sensitivity Prediction and Drug Sensitivity Biomarker Identification in Cancer

We present the Network-based Biased Tree Ensembles (NetBiTE) method for drug sensitivity prediction and drug sensitivity biomarker identification in cancer using a combination of prior knowledge and gene expression data. Our devised method consists of a biased tree ensemble that is built according to a probabilistic bias weight distribution. The bias weight distribution is obtained from the assignment of high weights to the drug targets and propagating the assigned weights over a protein-protein interaction network such as STRING. The propagation of weights, defines neighborhoods of influence around the drug targets and as such simulates the spread of perturbations within the cell, following drug administration. Using a synthetic dataset, we showcase how application of biased tree ensembles (BiTE) results in significant accuracy gains at a much lower computational cost compared to the unbiased random forests (RF) algorithm. We then apply NetBiTE to the Genomics of Drug Sensitivity in Cancer (GDSC) dataset and demonstrate that NetBiTE outperforms RF in predicting IC50 drug sensitivity, only for drugs that target membrane receptor pathways (MRPs): RTK, EGFR and IGFR signaling pathways. We propose based on the NetBiTE results, that for drugs that inhibit MRPs, the expression of target genes prior to drug administration is a biomarker for IC50 drug sensitivity following drug administration. We further verify and reinforce this proposition through control studies on, PI3K/MTOR signaling pathway inhibitors, a drug category that does not target MRPs, and through assignment of dummy targets to MRP inhibiting drugs and investigating the variation in NetBiTE accuracy.

FSP1 is a glutathione-independent ferroptosis suppressor

Ferroptosis is an iron-dependent form of necrotic cell death marked by oxidative damage to phospholipids1,2. To date, ferroptosis has been thought to be controlled only by the phospholipid hydroperoxide-reducing enzyme glutathione peroxidase 4 (GPX4)3,4 and radical-trapping antioxidants5,6. However, elucidation of the factors that underlie the sensitivity of a given cell type to ferroptosis7 is crucial to understand the pathophysiological role of ferroptosis and how it may be exploited for the treatment of cancer. Although metabolic constraints8 and phospholipid composition9,10 contribute to ferroptosis sensitivity, no cell-autonomous mechanisms have been identified that account for the resistance of cells to ferroptosis. Here we used an expression cloning approach to identify genes in human cancer cells that are able to complement the loss of GPX4. We found that the flavoprotein apoptosis-inducing factor mitochondria-associated 2 (AIFM2) is a previously unrecognized anti-ferroptotic gene. AIFM2, which we renamed ferroptosis suppressor protein 1 (FSP1) and which was initially described as a pro-apoptotic gene11, confers protection against ferroptosis elicited by GPX4 deletion. We further demonstrate that the suppression of ferroptosis by FSP1 is mediated by ubiquinone (also known as coenzyme Q10, CoQ10): the reduced form, ubiquinol, traps lipid peroxyl radicals that mediate lipid peroxidation, whereas FSP1 catalyses the regeneration of CoQ10 using NAD(P)H. Pharmacological targeting of FSP1 strongly synergizes with GPX4 inhibitors to trigger ferroptosis in a number of cancer entities. In conclusion, the FSP1-CoQ10-NAD(P)H pathway exists as a stand-alone parallel system, which co-operates with GPX4 and glutathione to suppress phospholipid peroxidation and ferroptosis.

Discovering long noncoding RNA predictors of anticancer drug sensitivity beyond protein-coding genes

Large-scale cancer cell line screens have identified thousands of protein-coding genes (PCGs) as biomarkers of anticancer drug response. However, systematic evaluation of long noncoding RNAs (lncRNAs) as pharmacogenomic biomarkers has so far proven challenging. Here, we study the contribution of lncRNAs as drug response predictors beyond spurious associations driven by correlations with proximal PCGs, tissue lineage, or established biomarkers. We show that, as a whole, the lncRNA transcriptome is equally potent as the PCG transcriptome at predicting response to hundreds of anticancer drugs. Analysis of individual lncRNAs transcripts associated with drug response reveals nearly half of the significant associations are in fact attributable to proximal cis-PCGs. However, adjusting for effects of cis-PCGs revealed significant lncRNAs that augment drug response predictions for most drugs, including those with well-established clinical biomarkers. In addition, we identify lncRNA-specific somatic alterations associated with drug response by adopting a statistical approach to determine lncRNAs carrying somatic mutations that undergo positive selection in cancer cells. Lastly, we experimentally demonstrate that 2 lncRNAs, EGFR-AS1 and MIR205HG, are functionally relevant predictors of anti-epidermal growth factor receptor (EGFR) drug response.

A Genome-wide Functional Signature Ontology Map and Applications to Natural Product Mechanism of Action Discovery

Gene expression signature-based inference of functional connectivity within and between genetic perturbations, chemical perturbations, and disease status can lead to the development of actionable hypotheses for gene function, chemical modes of action, and disease treatment strategies. Here, we report a FuSiOn-based genome-wide integration of hypomorphic cellular phenotypes that enables functional annotation of gene network topology, assignment of mechanistic hypotheses to genes of unknown function, and detection of cooperativity among cell regulatory systems. Dovetailing genetic perturbation data with chemical perturbation phenotypes allowed simultaneous generation of mechanism of action hypotheses for thousands of uncharacterized natural products fractions (NPFs). The predicted mechanism of actions span a broad spectrum of cellular mechanisms, many of which are not currently recognized as "druggable." To enable use of FuSiOn as a hypothesis generation resource, all associations and analyses are available within an open source web-based GUI (http://fusion.yuhs.ac).

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.

RWEN: response-weighted elastic net for prediction of chemosensitivity of cancer cell lines

Motivation

In recent years there have been several efforts to generate sensitivity profiles of collections of genomically characterized cell lines to panels of candidate therapeutic compounds. These data provide the basis for the development of in silico models of sensitivity based on cellular, genetic, or expression biomarkers of cancer cells. However, a remaining challenge is an efficient way to identify accurate sets of biomarkers to validate. To address this challenge, we developed methodology using gene-expression profiles of human cancer cell lines to predict the responses of these cell lines to a panel of compounds.

Results

We developed an iterative weighting scheme which, when applied to elastic net, a regularized regression method, significantly improves the overall accuracy of predictions, particularly in the highly sensitive response region. In addition to application of these methods to actual chemical sensitivity data, we investigated the effects of sample size, number of features, model sparsity, signal-to-noise ratio, and feature correlation on predictive performance using a simulation framework, particularly for situations where the number of covariates is much larger than sample size. While our method aims to be useful in therapeutic discovery and understanding of the basic mechanisms of action of drugs and their targets, it is generally applicable in any domain where predictions of extreme responses are of highest importance.

Availability and implementation

The iterative and other weighting algorithms were implemented in R. The code is available at https://github.com/kiwtir/RWEN. The CTRP data are available at ftp://caftpd.nci.nih.gov/pub/OCG-DCC/CTD2/Broad/CTRPv2.1_2016_pub_NatChemBiol_12_109/ and the Sanger data at ftp://ftp.sanger.ac.uk/pub/project/cancerrxgene/releases/release-6.0/.

Supplementary information

Supplementary data are available at Bioinformatics online.

Integration of Bioinformatics Resources Reveals the Therapeutic Benefits of Gemcitabine and Cell Cycle Intervention in SMAD4-Deleted Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is the most common and aggressive type of pancreatic cancer. The five-year survival rate of PDAC is very low (less than 8%), which is associated with the late diagnosis, high metastatic potential, and resistance to therapeutic agents. The identification of better prognostic or therapeutic biomarker may have clinical benefits for PDAC treatment. SMAD4, a central mediator of transforming growth factor beta (TGFβ) signaling pathway, is considered a tumor suppressor gene. SMAD4 inactivation is frequently found in PDAC. However, its role in prognosis and therapeutics of PDAC is still unclear. In this study, we applied bioinformatics approaches, and integrated publicly available resources, to investigate the role of SMAD4 gene deletion in PDAC. We found that SMAD4 deletion was associated with poorer disease-free, but not overall, survival in PDAC patients. Cancer hallmark enrichment and pathway analysis suggested that the upregulation of cell cycle-related genes in SMAD4-deleted PDAC. Chemotherapy response profiling of PDAC cell lines and patient-derived organoids revealed that SMAD4-deleted PDAC was sensitive to gemcitabine, the first-line treatment for PDAC, and specific cell cycle-targeting drugs. Taken together, our study provides an insight into the prognostic and therapeutic roles of SMAD4 gene deletion in PDAC, and SMAD4 gene copy numbers may be used as a therapeutic biomarker for PDAC treatment.

Modeling Cellular Response in Large-Scale Radiogenomic Databases to Advance Precision Radiotherapy

Radiotherapy is integral to the care of a majority of patients with cancer. Despite differences in tumor responses to radiation (radioresponse), dose prescriptions are not currently tailored to individual patients. Recent large-scale cancer cell line databases hold the promise of unravelling the complex molecular arrangements underlying cellular response to radiation, which is critical for novel predictive biomarker discovery. Here, we present RadioGx, a computational platform for integrative analyses of radioresponse using radiogenomic databases. We fit the dose-response data within RadioGx to the linear-quadratic model. The imputed survival across a range of dose levels (AUC) was a robust radioresponse indicator that correlated with biological processes known to underpin the cellular response to radiation. Using AUC as a metric for further investigations, we found that radiation sensitivity was significantly associated with disruptive mutations in genes related to nonhomologous end joining. Next, by simulating the effects of different oxygen levels, we identified putative genes that may influence radioresponse specifically under hypoxic conditions. Furthermore, using transcriptomic data, we found evidence for tissue-specific determinants of radioresponse, suggesting that tumor type could influence the validity of putative predictive biomarkers of radioresponse. Finally, integrating radioresponse with drug response data, we found that drug classes impacting the cytoskeleton, DNA replication, and mitosis display similar therapeutic effects to ionizing radiation on cancer cell lines. In summary, RadioGx provides a unique computational toolbox for hypothesis generation to advance preclinical research for radiation oncology and precision medicine. SIGNIFICANCE: The RadioGx computational platform enables integrative analyses of cellular response to radiation with drug responses and genome-wide molecular data. GRAPHICAL ABSTRACT: http://cancerres.aacrjournals.org/content/canres/79/24/6227/F1.large.jpg.See related commentary by Spratt and Speers, p. 6076.

Tankyrase disrupts metabolic homeostasis and promotes tumorigenesis by inhibiting LKB1-AMPK signalling

The LKB1/AMPK pathway plays a major role in cellular homeostasis and tumor suppression. Down-regulation of LKB1/AMPK occurs in several human cancers and has been implicated in metabolic diseases. However, the precise upstream regulation of LKB1-AMPK pathway is largely unknown. Here, we report that AMPK activation by LKB1 is regulated by tankyrases. Tankyrases interact with and ribosylate LKB1, promoting its K63-linked ubiquitination by an E3 ligase RNF146, which blocks LKB1/STRAD/MO25 complex formation and LKB1 activation. LKB1 activation by tankyrase inhibitors induces AMPK activation and suppresses tumorigenesis. Similarly, the tankyrase inhibitor G007-LK effectively regulates liver metabolism and glycemic control in diabetic mice in a LKB1-dependent manner. In patients with lung cancer, tankyrase levels negatively correlate with p-AMPK levels and poor survival. Taken together, these findings suggest that tankyrase and RNF146 are major up-stream regulators of LKB1-AMPK pathway and provide another focus for cancer and metabolic disease therapies.

Kernelized rank learning for personalized drug recommendation

Motivation

Large-scale screenings of cancer cell lines with detailed molecular profiles against libraries of pharmacological compounds are currently being performed in order to gain a better understanding of the genetic component of drug response and to enhance our ability to recommend therapies given a patient's molecular profile. These comprehensive screens differ from the clinical setting in which (i) medical records only contain the response of a patient to very few drugs, (ii) drugs are recommended by doctors based on their expert judgment and (iii) selecting the most promising therapy is often more important than accurately predicting the sensitivity to all potential drugs. Current regression models for drug sensitivity prediction fail to account for these three properties.

Results

We present a machine learning approach, named Kernelized Rank Learning (KRL), that ranks drugs based on their predicted effect per cell line (patient), circumventing the difficult problem of precisely predicting the sensitivity to the given drug. Our approach outperforms several state-of-the-art predictors in drug recommendation, particularly if the training dataset is sparse, and generalizes to patient data. Our work phrases personalized drug recommendation as a new type of machine learning problem with translational potential to the clinic.

Availability and implementation

The Python implementation of KRL and scripts for running our experiments are available at https://github.com/BorgwardtLab/Kernelized-Rank-Learning.

Supplementary information

Supplementary data are available at Bioinformatics online.

Structural characterization of EGFR exon 19 deletion mutation using molecular dynamics simulation

Epidermal growth factor receptor (EGFR) is a tyrosine kinase receptor important in diverse biological processes including cell proliferation and survival. Upregulation of EGFR activity due to over-expression or mutation is widely implicated in cancer. Activating somatic mutations of the EGFR kinase are postulated to affect the conformation and/or stability of the protein, shifting the EGFR inactive-active state equilibrium towards the activated state. Here, we examined a common EGFR deletion mutation, Δ⁷⁴⁶ELREA⁷⁵⁰, which is frequently observed in non-small cell lung cancer patients. By using molecular dynamics simulation, we investigated the structural effects of the mutation that lead to the experimentally reported increases in kinase activity. Simulations of the active form wild-type and ΔELREA EGFRs revealed the deletion stabilizes the αC helix of the kinase domain, which is located adjacent to the deletion site, by rigidifying the flexible β3-αC loop that accommodates the ELREA sequence. Consequently, the αC helix is stabilized in the “αC-in” active conformation that would prolong the time of the activated state. Moreover, in the mutant kinase, a salt bridge between E762 and K745, which is key for EGFR activity, was also stabilized during the simulation. Additionally, the interaction between EGFR and ATP was favored by ΔELREA EGFR over wild-type EGFR, as reflected by the number of hydrogen bonds formed and the free energy of binding. Simulation of inactive EGFR suggested the deletion would promote a shift from the inactive conformation towards active EGFR, which is supported by the inward movement of the αC helix. The MDS results also align with the effects of tyrosine kinase inhibitors on ΔELREA and wild-type EGFR lung cancer cell lines, where more pronounced inhibition was observed against ΔELREA than for wild-type EGFR by inhibitors recognizing the active kinase conformation.

In Vitro and In Silico Mechanistic Insights into miR-21-5p-Mediated Topoisomerase Drug Resistance in Human Colorectal Cancer Cells

Although chemotherapy for treating colorectal cancer has had some success, drug resistance and metastasis remain the major causes of death for colorectal cancer patients. MicroRNA-21-5p (hereafter denoted as miR-21) is one of the most abundant miRNAs in human colorectal cancer. A Kaplan-Meier survival analysis found a negative prognostic correlation of miR-21 and metastasis-free survival in colorectal cancer patients (The Cancer Genome Atlas Colon Adenocarcinoma/TCGA-COAD cohort). To explore the role of miR-21 overexpression in drug resistance, a stable miR-21-overexpressing clone in a human DLD-1 colorectal cancer cell line was established. The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) cell viability assay found that miR-21 overexpression induced drug resistance to topoisomerase inhibitors (SN-38, doxorubicin, and etoposide/VP-16). Mechanistically, we showed that miR-21 overexpression reduced VP-16-induced apoptosis and concomitantly enhanced pro-survival autophagic flux without the alteration of topoisomerase expression and activity. Bioinformatics analyses suggested that miR-21 overexpression induced genetic reprogramming that mimicked the gene signature of topoisomerase inhibitors and downregulated genes related to the proteasome pathway. Taken together, our results provide a novel insight into the role of miR-21 in the development of drug resistance in colorectal cancer.

Looking beyond the hype: Applied AI and machine learning in translational medicine

Big data problems are becoming more prevalent for laboratory scientists who look to make clinical impact. A large part of this is due to increased computing power, in parallel with new technologies for high quality data generation. Both new and old techniques of artificial intelligence (AI) and machine learning (ML) can now help increase the success of translational studies in three areas: drug discovery, imaging, and genomic medicine. However, ML technologies do not come without their limitations and shortcomings. Current technical limitations and other limitations including governance, reproducibility, and interpretation will be discussed in this article. Overcoming these limitations will enable ML methods to be more powerful for discovery and reduce ambiguity within translational medicine, allowing data-informed decision-making to deliver the next generation of diagnostics and therapeutics to patients quicker, at lowered costs, and at scale.

In Vivo ERK1/2 Reporter Predictively Models Response and Resistance to Combined BRAF and MEK Inhibitors in Melanoma

Combined BRAF and MEK inhibition is a standard of care in patients with advanced BRAF-mutant melanoma, but acquired resistance remains a challenge that limits response durability. Here, we quantitated in vivo ERK1/2 activity and tumor response associated with resistance to combined BRAF and MEK inhibition in mutant BRAF xenografts. We found that ERK1/2 pathway reactivation preceded the growth of resistant tumors. Moreover, we detected a subset of cells that not only persisted throughout long-term treatment but restored ERK1/2 signaling and grew upon drug removal. Cell lines derived from combination-resistant tumors (CRT) exhibited elevated ERK1/2 phosphorylation, which were sensitive to ERK1/2 inhibition. In some CRTs, we detected a tandem duplication of the BRAF kinase domain. Monitoring ERK1/2 activity in vivo was efficacious in predicting tumor response during intermittent treatment. We observed maintained expression of the mitotic regulator, polo-like kinase 1 (Plk1), in melanoma resistant to BRAF and MEK inhibitors. Plk1 inhibition induced apoptosis in CRTs, leading to slowed growth of BRAF and MEK inhibitor-resistant tumors in vivo These data demonstrate the utility of in vivo ERK1/2 pathway reporting as a tool to optimize clinical dosing schemes and establish suppression of Plk1 as potential salvage therapy for BRAF inhibitor and MEK inhibitor-resistant melanoma.

Co-occurring genomic alterations in non-small-cell lung cancer biology and therapy

The impressive clinical activity of small-molecule receptor tyrosine kinase inhibitors for oncogene-addicted subgroups of non-small-cell lung cancer (for example, those driven by activating mutations in the gene encoding epidermal growth factor receptor (EGFR) or rearrangements in the genes encoding the receptor tyrosine kinases anaplastic lymphoma kinase (ALK), ROS proto-oncogene 1 (ROS1) and rearranged during transfection (RET)) has established an oncogene-centric molecular classification paradigm in this disease. However, recent studies have revealed considerable phenotypic diversity downstream of tumour-initiating oncogenes. Co-occurring genomic alterations, particularly in tumour suppressor genes such as TP53 and LKB1 (also known as STK11), have emerged as core determinants of the molecular and clinical heterogeneity of oncogene-driven lung cancer subgroups through their effects on both tumour cell-intrinsic and non-cell-autonomous cancer hallmarks. In this Review, we discuss the impact of co-mutations on the pathogenesis, biology, microenvironmental interactions and therapeutic vulnerabilities of non-small-cell lung cancer and assess the challenges and opportunities that co-mutations present for personalized anticancer therapy, as well as the expanding field of precision immunotherapy.

Expression Levels of Therapeutic Targets as Indicators of Sensitivity to Targeted Therapeutics

Cancer precision medicine aims to predict the drug likely to yield the best response for a patient. Genomic sequencing of tumors is currently being used to better inform treatment options; however, this approach has had a limited clinical impact due to the paucity of actionable mutations. An alternative to mutation status is the use of gene expression signatures to predict response. Using data from two large-scale studies, The Genomics of Drug Sensitivity of Cancer (GDSC) and The Cancer Therapeutics Response Portal (CTRP), we investigated the relationship between the sensitivity of hundreds of cell lines to hundreds of drugs, and the relative expression levels of the targets these drugs are directed against. For approximately one third of the drugs considered (73/222 in GDSC and 131/360 in CTRP), sensitivity was significantly correlated with the expression of at least one of the known targets. Surprisingly, for 8% of the annotated targets, there was a significant anticorrelation between target expression and sensitivity. For several cases, this corresponded to drugs targeting multiple genes in the same family, with the expression of one target significantly correlated with sensitivity and another significantly anticorrelated suggesting a possible role in resistance. Furthermore, we identified nontarget genes that are significantly correlated or anticorrelated with drug sensitivity, and find literature linking several to sensitization and resistance. Our analyses provide novel and important insights into both potential mechanisms of resistance and relative efficacy of drugs against the same target.

A Pharmacogenomic Landscape in Human Liver Cancers

Liver cancers are highly heterogeneous with poor prognosis and drug response. A better understanding between genetic alterations and drug responses would facilitate precision treatment for liver cancers. To characterize the landscape of pharmacogenomic interactions in liver cancers, we developed a protocol to establish human liver cancer cell models at a success rate of around 50% and generated the Liver Cancer Model Repository (LIMORE) with 81 cell models. LIMORE represented genomic and transcriptomic heterogeneity of primary cancers. Interrogation of the pharmacogenomic landscape of LIMORE discovered unexplored gene-drug associations, including synthetic lethalities to prevalent alterations in liver cancers. Moreover, predictive biomarker candidates were suggested for the selection of sorafenib-responding patients. LIMORE provides a rich resource facilitating drug discovery in liver cancers.

GDA, a web-based tool for Genomics and Drugs integrated analysis

Several major screenings of genetic profiling and drug testing in cancer cell lines proved that the integration of genomic portraits and compound activities is effective in discovering new genetic markers of drug sensitivity and clinically relevant anticancer compounds. Despite most genetic and drug response data are publicly available, the availability of user-friendly tools for their integrative analysis remains limited, thus hampering an effective exploitation of this information. Here, we present GDA, a web-based tool for Genomics and Drugs integrated Analysis that combines drug response data for >50 800 compounds with mutations and gene expression profiles across 73 cancer cell lines. Genomic and pharmacological data are integrated through a modular architecture that allows users to identify compounds active towards cancer cell lines bearing a specific genomic background and, conversely, the mutational or transcriptional status of cells responding or not-responding to a specific compound. Results are presented through intuitive graphical representations and supplemented with information obtained from public repositories. As both personalized targeted therapies and drug-repurposing are gaining increasing attention, GDA represents a resource to formulate hypotheses on the interplay between genomic traits and drug response in cancer. GDA is freely available at http://gda.unimore.it/.

Patient-Derived Xenografts and Matched Cell Lines Identify Pharmacogenomic Vulnerabilities in Colorectal Cancer

PURPOSE

Patient-derived xenograft (PDX) models accurately recapitulate the tumor of origin in terms of histopathology, genomic landscape, and therapeutic response, but some limitations due to costs associated with their maintenance and restricted amenability for large-scale screenings still exist. To overcome these issues, we established a platform of 2D cell lines (xeno-cell lines, XL), derived from PDXs of colorectal cancer with matched patient germline gDNA available.

EXPERIMENTAL DESIGN

Whole-exome and transcriptome sequencing analyses were performed. Biomarkers of response and resistance to anti-HER therapy were annotated. Dependency on the WRN helicase gene was assessed in MSS, MSI-H, and MSI-like XLs using a reverse genetics functional approach.

RESULTS

XLs recapitulated the entire spectrum of colorectal cancer transcriptional subtypes. Exome and RNA-seq analyses delineated several molecular biomarkers of response and resistance to EGFR and HER2 blockade. Genotype-driven responses observed in vitro in XLs were confirmed in vivo in the matched PDXs. MSI-H models were dependent upon WRN gene expression, while loss of WRN did not affect MSS XLs growth. Interestingly, one MSS XL with transcriptional MSI-like traits was sensitive to WRN depletion.

CONCLUSIONS

The XL platform represents a preclinical tool for functional gene validation and proof-of-concept studies to identify novel druggable vulnerabilities in colorectal cancer.

Molecular characteristics and therapeutic vulnerabilities across paediatric solid tumours

The spectrum of tumours arising in childhood is fundamentally different from that seen in adults, and they are known to be divergent from adult malignancies in terms of cellular origins, epidemiology, genetic complexity, driver mutations and underlying mutational processes. Despite the immense knowledge generated through sequencing efforts and functional characterization of identified (epi-)genetic alterations over the past decade, the clinical implications of this knowledge have so far been limited. Novel preclinical platforms such as the European Innovative Therapies for Children with Cancer-Paediatric Preclinical Proof-of-Concept Platform and the US-based Pediatric Preclinical Testing Consortium are being developed to try to change this by aiming to recapitulate the extensive heterogeneity of paediatric tumours and thereby, hopefully, improve the ability to predict clinical benefit. Numerous studies have also been established worldwide to provide patients with access to real-time molecular profiling and the possibility of more precise mechanism-of-action-based treatments. In addition to tumour-intrinsic findings and mechanisms, ongoing studies are investigating features such as the immune microenvironment of paediatric tumours in comparison with adult cancers - currently of very timely clinical relevance. However, there is an ongoing need for rigorous preclinical biomarker and target validation to feed into the next generation of molecularly stratified clinical trials. This Review aims to provide a comprehensive state-of-the-art overview of the molecular landscape of paediatric solid tumours, including their underlying genomic alterations and interactions with the microenvironment, complemented with our current understanding of potential therapeutic vulnerabilities and how these can be preclinically tested using more accurate predictive methods. Finally, we provide an outlook on the challenges and opportunities associated with translating this overwhelming scientific progress into real clinical benefit.

PharmacoDB: an integrative database for mining in vitro anticancer drug screening studies

Recent cancer pharmacogenomic studies profiled large panels of cell lines against hundreds of approved drugs and experimental chemical compounds. The overarching goal of these screens is to measure sensitivity of cell lines to chemical perturbations, correlate these measures to genomic features, and thereby develop novel predictors of drug response. However, leveraging these valuable data is challenging due to the lack of standards for annotating cell lines and chemical compounds, and quantifying drug response. Moreover, it has been recently shown that the complexity and complementarity of the experimental protocols used in the field result in high levels of technical and biological variation in the in vitro pharmacological profiles. There is therefore a need for new tools to facilitate rigorous comparison and integrative analysis of large-scale drug screening datasets. To address this issue, we have developed PharmacoDB (pharmacodb.pmgenomics.ca), a database integrating the largest cancer pharmacogenomic studies published to date. Here, we describe how the curation of cell line and chemical compound identifiers maximizes the overlap between datasets and how users can leverage such data to compare and extract robust drug phenotypes. PharmacoDB provides a unique resource to mine a compendium of curated cancer pharmacogenomic datasets that are otherwise disparate and difficult to integrate.

Decomposing Oncogenic Transcriptional Signatures to Generate Maps of Divergent Cellular States

The systematic sequencing of the cancer genome has led to the identification of numerous genetic alterations in cancer. However, a deeper understanding of the functional consequences of these alterations is necessary to guide appropriate therapeutic strategies. Here, we describe Onco-GPS (OncoGenic Positioning System), a data-driven analysis framework to organize individual tumor samples with shared oncogenic alterations onto a reference map defined by their underlying cellular states. We applied the methodology to the RAS pathway and identified nine distinct components that reflect transcriptional activities downstream of RAS and defined several functional states associated with patterns of transcriptional component activation that associates with genomic hallmarks and response to genetic and pharmacological perturbations. These results show that the Onco-GPS is an effective approach to explore the complex landscape of oncogenic cellular states across cancers, and an analytic framework to summarize knowledge, establish relationships, and generate more effective disease models for research or as part of individualized precision medicine paradigms.

Screening of Drug Repositioning Candidates for Castration Resistant Prostate Cancer

Purpose: Most prostate cancers (PCs) initially respond to androgen deprivation therapy (ADT), but eventually many PC patients develop castration resistant PC (CRPC). Currently, available drugs that have been approved for the treatment of CRPC patients are limited. Computational drug repositioning methods using public databases represent a promising and efficient tool for discovering new uses for existing drugs. The purpose of the present study is to predict drug candidates that can treat CRPC using a computational method that integrates publicly available gene expression data of tumors from CRPC patients, drug-induced gene expression data and drug response activity data. Methods: Gene expression data from tumoral and normal or benign prostate tissue samples in CRPC patients were downloaded from the Gene Expression Omnibus (GEO) and differentially expressed genes (DEGs) in CRPC were determined with a meta-signature analysis by a metaDE R package. Additionally, drug activity data were downloaded from the ChEMBL database. Furthermore, the drug-induced gene expression data were downloaded from the LINCS database. The reversal relationship between the CRPC and drug gene expression signatures as the Reverse Gene Expression Scores (RGES) were computed. Drug candidates to treat CRPC were predicted using summarized scores (sRGES). Additionally, synergic effects of drug combinations were predicted with a Target Inhibition interaction using the Minimization and Maximization Averaging (TIMMA) algorithm. Results: The drug candidates of sorafenib, olaparib, elesclomol, tanespimycin, and ponatinib were predicted to be active for the treatment of CRPC. Meanwhile, CRPC-related genes, in this case MYL9, E2F2, APOE, and ZFP36, were identified as having gene expression data that can be reversed by these drugs. Additionally, lenalidomide in combination with pazopanib was predicted to be most potent for CRPC. Conclusion: These findings support the use of a computational reversal gene expression approach to identify new drug and drug combination candidates that can be used to treat CRPC.

A Multi-center Study on the Reproducibility of Drug-Response Assays in Mammalian Cell Lines

Evidence that some high-impact biomedical results cannot be repeated has stimulated interest in practices that generate findable, accessible, interoperable, and reusable (FAIR) data. Multiple papers have identified specific examples of irreproducibility, but practical ways to make data more reproducible have not been widely studied. Here, five research centers in the NIH LINCS Program Consortium investigate the reproducibility of a prototypical perturbational assay: quantifying the responsiveness of cultured cells to anti-cancer drugs. Such assays are important for drug development, studying cellular networks, and patient stratification. While many experimental and computational factors impact intra- and inter-center reproducibility, the factors most difficult to identify and control are those with a strong dependency on biological context. These factors often vary in magnitude with the drug being analyzed and with growth conditions. We provide ways to identify such context-sensitive factors, thereby improving both the theory and practice of reproducible cell-based assays.

Factors that impact the reproducibility of experimental data are poorly understood. Five NIH-LINCS centers performed the same set of drug-response measurements and compared results. Technical and biological variables that impact precision and reproducibility and are also sensitive to biological context were the most problematic.

mTCTScan: a comprehensive platform for annotation and prioritization of mutations affecting drug sensitivity in cancers

Cancer therapies have experienced rapid progress in recent years, with a number of novel small-molecule kinase inhibitors and monoclonal antibodies now being widely used to treat various types of human cancers. During cancer treatments, mutations can have important effects on drug sensitivity. However, the relationship between tumor genomic profiles and the effectiveness of cancer drugs remains elusive. We introduce Mutation To Cancer Therapy Scan (mTCTScan) web server (http://jjwanglab.org/mTCTScan) that can systematically analyze mutations affecting cancer drug sensitivity based on individual genomic profiles. The platform was developed by leveraging the latest knowledge on mutation-cancer drug sensitivity associations and the results from large-scale chemical screening using human cancer cell lines. Using an evidence-based scoring scheme based on current integrative evidences, mTCTScan is able to prioritize mutations according to their associations with cancer drugs and preclinical compounds. It can also show related drugs/compounds with sensitivity classification by considering the context of the entire genomic profile. In addition, mTCTScan incorporates comprehensive filtering functions and cancer-related annotations to better interpret mutation effects and their association with cancer drugs. This platform will greatly benefit both researchers and clinicians for interrogating mechanisms of mutation-dependent drug response, which will have a significant impact on cancer precision medicine.

The cancer cell proteome and transcriptome predicts sensitivity to targeted and cytotoxic drugs

This study shows that the proteomic and transcriptomic states of cancer cells are more predictive of drug sensitivity than genomic markers for most drugs, both within and across tumor types.

Tumors of different molecular subtypes can show strongly deviating responses to drug treatment, making stratification of patients based on molecular markers an important part of cancer therapy. Pharmacogenomic studies have led to the discovery of selected genomic markers (e.g., BRAF^V600E), whereas transcriptomic and proteomic markers so far have been largely absent in clinical use, thus constituting a potentially valuable resource for further substratification of patients. To systematically assess the explanatory power of different -omics data types, we assembled a panel of 49 melanoma cell lines, including genomic, transcriptomic, proteomic, and pharmacological data, showing that drug sensitivity models trained on transcriptomic or proteomic data outperform genomic-based models for most drugs. These results were confirmed in eight additional tumor types using published datasets. Furthermore, we show that drug sensitivity models can be transferred between tumor types, although after correcting for training sample size, transferred models perform worse than within-tumor–type predictions. Our results suggest that transcriptomic/proteomic signals may be alternative biomarker candidates for the stratification of patients without known genomic markers.

KekuleScope: prediction of cancer cell line sensitivity and compound potency using convolutional neural networks trained on compound images

The application of convolutional neural networks (ConvNets) to harness high-content screening images or 2D compound representations is gaining increasing attention in drug discovery. However, existing applications often require large data sets for training, or sophisticated pretraining schemes. Here, we show using 33 IC50 data sets from ChEMBL 23 that the in vitro activity of compounds on cancer cell lines and protein targets can be accurately predicted on a continuous scale from their Kekulé structure representations alone by extending existing architectures (AlexNet, DenseNet-201, ResNet152 and VGG-19), which were pretrained on unrelated image data sets. We show that the predictive power of the generated models, which just require standard 2D compound representations as input, is comparable to that of Random Forest (RF) models and fully-connected Deep Neural Networks trained on circular (Morgan) fingerprints. Notably, including additional fully-connected layers further increases the predictive power of the ConvNets by up to 10%. Analysis of the predictions generated by RF models and ConvNets shows that by simply averaging the output of the RF models and ConvNets we obtain significantly lower errors in prediction for multiple data sets, although the effect size is small, than those obtained with either model alone, indicating that the features extracted by the convolutional layers of the ConvNets provide complementary predictive signal to Morgan fingerprints. Lastly, we show that multi-task ConvNets trained on compound images permit to model COX isoform selectivity on a continuous scale with errors in prediction comparable to the uncertainty of the data. Overall, in this work we present a set of ConvNet architectures for the prediction of compound activity from their Kekulé structure representations with state-of-the-art performance, that require no generation of compound descriptors or use of sophisticated image processing techniques. The code needed to reproduce the results presented in this study and all the data sets are provided at https://github.com/isidroc/kekulescope .

Why imaging data alone is not enough: AI-based integration of imaging, omics, and clinical data

Artificial intelligence (AI) is currently regaining enormous interest due to the success of machine learning (ML), and in particular deep learning (DL). Image analysis, and thus radiomics, strongly benefits from this research. However, effectively and efficiently integrating diverse clinical, imaging, and molecular profile data is necessary to understand complex diseases, and to achieve accurate diagnosis in order to provide the best possible treatment. In addition to the need for sufficient computing resources, suitable algorithms, models, and data infrastructure, three important aspects are often neglected: (1) the need for multiple independent, sufficiently large and, above all, high-quality data sets; (2) the need for domain knowledge and ontologies; and (3) the requirement for multiple networks that provide relevant relationships among biological entities. While one will always get results out of high-dimensional data, all three aspects are essential to provide robust training and validation of ML models, to provide explainable hypotheses and results, and to achieve the necessary trust in AI and confidence for clinical applications.

Big data in pharmacogenomics: current applications, perspectives and pitfalls

The efficiency of new generation sequencing methods and the reduction of their cost has led pharmacogenomics to gradually supplant pharmacogenetics, leading to new applications in personalized medicine along with new perspectives in drug design or identification of drug response factors. The amount of data generated in genomics fits the definition of big data, and need a specific bioinformatics processing following standard steps: data collection, processing, analysis and interpretation. Pitfalls of pharmacogenomics studies are directly related to these steps. This review aims to describe these steps from a pharmacogenomic point of view, focusing on bioinformatics aspects.

Identifying Cancer Driver lncRNAs Bridged by Functional Effectors through Integrating Multi-omics Data in Human Cancers

The accumulation of somatic driver mutations in the human genome enables cells to gradually acquire a growth advantage and contributes to tumor development. Great efforts on protein-coding cancer drivers have yielded fruitful discoveries and clinical applications. However, investigations on cancer drivers in non-coding regions, especially long non-coding RNAs (lncRNAs), are extremely scarce due to the limitation of functional understanding. Thus, to identify driver lncRNAs integrating multi-omics data in human cancers, we proposed a computational framework, DriverLncNet, which dissected the functional impact of somatic copy number alteration (CNA) of lncRNAs on regulatory networks and captured key functional effectors in dys-regulatory networks. Applying it to 5 cancer types from The Cancer Genome Atlas (TCGA), we portrayed the landscape of 117 driver lncRNAs and revealed their associated cancer hallmarks through their functional effectors. Moreover, lncRNA RP11-571M6.8 was detected to be highly associated with immunotherapeutic targets (PD-1, PD-L1, and CTLA-4) and regulatory T cell infiltration level and their markers (IL2RA and FCGR2B) in glioblastoma multiforme, highlighting its immunosuppressive function. Meanwhile, a high expression of RP11-1020A11.1 in bladder carcinoma was predictive of poor survival independent of clinical characteristics, and CTD-2256P15.2 in lung adenocarcinoma responded to the sensitivity of methyl ethyl ketone (MEK) inhibitors. In summary, this study provided a framework to decipher the mechanisms of tumorigenesis from driver lncRNA level, established a new landscape of driver lncRNAs in human cancers, and offered potential clinical implications for precision oncology.

The ETS Inhibitors YK-4-279 and TK-216 Are Novel Antilymphoma Agents

PURPOSE

Transcription factors are commonly deregulated in cancer, and they have been widely considered as difficult to target due to their nonenzymatic mechanism of action. Altered expression levels of members of the ETS-transcription factors are often observed in many different tumors, including lymphomas. Here, we characterized two small molecules, YK-4-279 and its clinical derivative, TK-216, targeting ETS factors via blocking the protein-protein interaction with RNA helicases, for their antilymphoma activity.

EXPERIMENTAL DESIGN

The study included preclinical in vitro activity screening on a large panel of cell lines, both as single agent and in combination; validation experiments on in vivo models; and transcriptome and coimmunoprecipitation experiments.

RESULTS

YK-4-279 and TK-216 demonstrated an antitumor activity across several lymphoma cell lines, which we validated in vivo. We observed synergistic activity when YK-4-279 and TK-216 were combined with the BCL2 inhibitor venetoclax and with the immunomodulatory drug lenalidomide. YK-4-279 and TK-216 interfere with protein interactions of ETS family members SPIB, in activated B-cell-like type diffuse large B-cell lymphomas, and SPI1, in germinal center B-cell-type diffuse large B-cell lymphomas.

CONCLUSIONS

The ETS inhibitor YK-4-279 and its clinical derivative TK-216 represent a new class of agents with in vitro and in vivo antitumor activity in lymphomas. Although their detailed mechanism of action needs to be fully defined, in DLBCL they might act by targeting subtype-specific essential transcription factors.

Anticancer Drug Response Prediction in Cell Lines Using Weighted Graph Regularized Matrix Factorization

Precision medicine has become a novel and rising concept, which depends much on the identification of individual genomic signatures for different patients. The cancer cell lines could reflect the “omic” diversity of primary tumors, based on which many works have been carried out to study the cancer biology and drug discovery both in experimental and computational aspects. In this work, we presented a novel method to utilize weighted graph regularized matrix factorization (WGRMF) for inferring anticancer drug response in cell lines. We constructed a p-nearest neighbor graph to sparsify drug similarity matrix and cell line similarity matrix, respectively. Using the sparsified matrices in the graph regularization terms, we performed matrix factorization to generate the latent matrices for drug and cell line. The graph regularization terms including neighbor information could help to exclude the noisy ingredient and improve the prediction accuracy. The 10-fold cross-validation was implemented, and the Pearson correlation coefficient (PCC), root-mean-square error (RMSE), PCCsr, and RMSEsr averaged over all drugs were calculated to evaluate the performance of WGRMF. The results on the Genomics of Drug Sensitivity in Cancer (GDSC) dataset are 0.64 ± 0.16, 1.37 ± 0.35, 0.73 ± 0.14, and 1.71 ± 0.44 for PCC, RMSE, PCCsr, and RMSEsr in turn. And for the Cancer Cell Line Encyclopedia (CCLE) dataset, WGRMF got results of 0.72 ± 0.09, 0.56 ± 0.19, 0.79 ± 0.07, and 0.69 ± 0.19, respectively. The results showed the superiority of WGRMF compared with previous methods. Besides, based on the prediction results using the GDSC dataset, three types of case studies were carried out. The results from both cross-validation and case studies have shown the effectiveness of WGRMF on the prediction of drug response in cell lines.

Neuronal differentiation and cell-cycle programs mediate response to BET-bromodomain inhibition in MYC-driven medulloblastoma

BET-bromodomain inhibition (BETi) has shown pre-clinical promise for MYC-amplified medulloblastoma. However, the mechanisms for its action, and ultimately for resistance, have not been fully defined. Here, using a combination of expression profiling, genome-scale CRISPR/Cas9-mediated loss of function and ORF/cDNA driven rescue screens, and cell-based models of spontaneous resistance, we identify bHLH/homeobox transcription factors and cell-cycle regulators as key genes mediating BETi's response and resistance. Cells that acquire drug tolerance exhibit a more neuronally differentiated cell-state and expression of lineage-specific bHLH/homeobox transcription factors. However, they do not terminally differentiate, maintain expression of CCND2, and continue to cycle through S-phase. Moreover, CDK4/CDK6 inhibition delays acquisition of resistance. Therefore, our data provide insights about the mechanisms underlying BETi effects and the appearance of resistance and support the therapeutic use of combined cell-cycle inhibitors with BETi in MYC-amplified medulloblastoma.

Sparse discriminative latent characteristics for predicting cancer drug sensitivity from genomic features

Drug screening studies typically involve assaying the sensitivity of a range of cancer cell lines across an array of anti-cancer therapeutics. Alongside these sensitivity measurements high dimensional molecular characterizations of the cell lines are typically available, including gene expression, copy number variation and genomic mutations. We propose a sparse multitask regression model which learns discriminative latent characteristics that predict drug sensitivity and are associated with specific molecular features. We use ideas from Bayesian nonparametrics to automatically infer the appropriate number of these latent characteristics. The resulting analysis couples high predictive performance with interpretability since each latent characteristic involves a typically small set of drugs, cell lines and genomic features. Our model uncovers a number of drug-gene sensitivity associations missed by single gene analyses. We functionally validate one such novel association: that increased expression of the cell-cycle regulator C/EBPδ decreases sensitivity to the histone deacetylase (HDAC) inhibitor panobinostat.

Non-canonical cMet regulation by vimentin mediates Plk1 inhibitor-induced apoptosis

To address the need for improved systemic therapy for non-small-cell lung cancer (NSCLC), we previously demonstrated that mesenchymal NSCLC was sensitive to polo-like kinase (Plk1) inhibitors, but the mechanisms of resistance in epithelial NSCLC remain unknown. Here, we show that cMet was differentially regulated in isogenic pairs of epithelial and mesenchymal cell lines. Plk1 inhibition inhibits cMet phosphorylation only in mesenchymal cells. Constitutively active cMet abrogates Plk1 inhibitor-induced apoptosis. Likewise, cMet silencing or inhibition enhances Plk1 inhibitor-induced apoptosis. Cells with acquired resistance to Plk1 inhibitors are more epithelial than their parental cells and maintain cMet activation after Plk1 inhibition. In four animal NSCLC models, mesenchymal tumors were more sensitive to Plk1 inhibition alone than were epithelial tumors. The combination of cMet and Plk1 inhibition led to regression of tumors that did not regrow when drug treatment was stopped. Plk1 inhibition did not affect HGF levels but did decrease vimentin phosphorylation, which regulates cMet phosphorylation via β1-integrin. This research defines a heretofore unknown mechanism of ligand-independent activation of cMet downstream of Plk1 and an effective combination therapy.

A GPX4-dependent cancer cell state underlies the clear-cell morphology and confers sensitivity to ferroptosis

Clear-cell carcinomas (CCCs) are a histological group of highly aggressive malignancies commonly originating in the kidney and ovary. CCCs are distinguished by aberrant lipid and glycogen accumulation and are refractory to a broad range of anti-cancer therapies. Here we identify an intrinsic vulnerability to ferroptosis associated with the unique metabolic state in CCCs. This vulnerability transcends lineage and genetic landscape, and can be exploited by inhibiting glutathione peroxidase 4 (GPX4) with small-molecules. Using CRISPR screening and lipidomic profiling, we identify the hypoxia-inducible factor (HIF) pathway as a driver of this vulnerability. In renal CCCs, HIF-2α selectively enriches polyunsaturated lipids, the rate-limiting substrates for lipid peroxidation, by activating the expression of hypoxia-inducible, lipid droplet-associated protein (HILPDA). Our study suggests targeting GPX4 as a therapeutic opportunity in CCCs, and highlights that therapeutic approaches can be identified on the basis of cell states manifested by morphological and metabolic features in hard-to-treat cancers.

CDRgator: An Integrative Navigator of Cancer Drug Resistance Gene Signatures

Understanding the mechanisms of cancer drug resistance is a critical challenge in cancer therapy. For many cancer drugs, various resistance mechanisms have been identified such as target alteration, alternative signaling pathways, epithelial-mesenchymal transition, and epigenetic modulation. Resistance may arise via multiple mechanisms even for a single drug, making it necessary to investigate multiple independent models for comprehensive understanding and therapeutic application. In particular, we hypothesize that different resistance processes result in distinct gene expression changes. Here, we present a web-based database, CDRgator (Cancer Drug Resistance navigator) for comparative analysis of gene expression signatures of cancer drug resistance. Resistance signatures were extracted from two different types of datasets. First, resistance signatures were extracted from transcriptomic profiles of cancer cells or patient samples and their resistance-induced counterparts for >30 cancer drugs. Second, drug resistance group signatures were also extracted from two large-scale drug sensitivity datasets representing ~1,000 cancer cell lines. All the datasets are available for download, and are conveniently accessible based on drug class and cancer type, along with analytic features such as clustering analysis, multidimensional scaling, and pathway analysis. CDRgator allows meta-analysis of independent resistance models for more comprehensive understanding of drug-resistance mechanisms that is difficult to accomplish with individual datasets alone (database URL: http://cdrgator.ewha.ac.kr).