A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics

Danh-Tai Hoang, Gal Dinstag, Eldad D. Shulman, Leandro C. Hermida, Doreen S. Ben-Zvi, Efrat Elis, Katherine Caley, Stephen-John Sammut, Sanju Sinha, Neelam Sinha, Christopher H. Dampier, Chani Stossel, Tejas Patil, Arun Rajan, Wiem Lassoued, Julius Strauss, Shania Bailey, Clint Allen, Jason Redman, Tuvik Beker, Peng Jiang, Talia Golan, Scott Wilkinson, Adam G. Sowalsky, Sharon R. Pine, Carlos Caldas, James L. Gulley, Kenneth Aldape, Ranit Aharonov, Eric A. Stone & Eytan Ruppin

July, 2024

Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin-stained tumor slides for precision oncology. We present ENLIGHT–DeepPT, an indirect two-step approach consisting of DeepPT, a deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and ENLIGHT, which predicts response to targeted and immune therapies from the inferred expression values. We show that DeepPT successfully predicts transcriptomics in all 16 The Cancer Genome Atlas cohorts tested and generalizes well to two independent datasets. ENLIGHT–DeepPT successfully predicts true responders in five independent patient cohorts involving four different treatments spanning six cancer types, with an overall odds ratio of 2.28 and a 39.5% increased response rate among predicted responders versus the baseline rate. Notably, its prediction accuracy, obtained without any training on the treatment data, is comparable to that achieved by directly predicting the response from the images, which requires specific training on the treatment evaluation cohorts.

 

Read the full publication on Nature Cancer

Spectrum of Response to Platinum and PARP Inhibitors in Germline BRCA–Associated Pancreatic Cancer in the Clinical and Preclinical Setting

Chani Stossel,  Maria Raitses-Gurevich, Dikla Atias, Tamar Beller, Yulia Glick Gorman, Sharon Halperin, Eyal Peer, Robert E. Denroche, Amy Zhang, Faiyaz Notta, Julie M. Wilson, Grainne M. O’Kane, Elina Haimov Talmoud, Nora Amison, Michael Schvimer, Seth J. Salpeter, Vered Bar, Adi Zundelevich, Itay Tirosh, Rotem Tal, Gal Dinstag, Yaron Kinar,  Yonatan Eliezer, Uri Ben-David, Nancy S. Gavert, Ravid Straussman, Steven J. Gallinger, Raanan Berger, Talia Golan.
August, 2023

Germline BRCA–associated pancreatic ductal adenocarcinoma (glBRCA PDAC) tumors are susceptible to platinum and PARP inhibition. The clinical outcomes of 125 patients with glBRCA PDAC were stratified based on the spectrum of response to platinum/PARP inhibition: (i) refractory [overall survival (OS) <6 months], (ii) durable response followed by acquired resistance (OS <36 months), and (iii) long-term responders (OS >36 months). Patient-derived xenografts (PDX) were generated from 25 patients with glBRCA PDAC at different clinical time points. Response to platinum/PARP inhibition in vivo and ex vivo culture (EVOC) correlated with clinical response. We deciphered the mechanisms of resistance in glBRCA PDAC and identified homologous recombination (HR) proficiency and secondary mutations restoring partial functionality as the most dominant resistant mechanism. Yet, a subset of HR-deficient (HRD) patients demonstrated clinical resistance. Their tumors displayed basal-like molecular subtype and were more aneuploid. Tumor mutational burden was high in HRD PDAC and significantly higher in tumors with secondary mutations. Anti–PD-1 attenuated tumor growth in a novel humanized glBRCA PDAC PDX model. This work demonstrates the utility of preclinical models, including EVOC, to predict the response of glBRCA PDAC to treatment, which has the potential to inform time-sensitive medical decisions.

 

Read the full publication here

Clinically oriented prediction of patient response to targeted and immunotherapies from the tumor transcriptome

Gal Dinstag, Eldad D. Shulman, Efrat Elis, Doreen S. Ben-Zvi, Omer Tirosh, Eden Maimon, Isaac Meilijson, Emmanuel Elalouf, Boris Temkin, Philipp Vitkovsky, Eyal Schiff, Danh-Tai Hoang, Sanju Sinha, Nishanth Ulhas Nair, Joo Sang Lee, Alejandro A. Schäffer, Ze’ev Ronai, Dejan Juric, Andrea B. Apolo, William L. Dahut, Stanley Lipkowitz, Raanan Berger, Razelle Kurzrock, Antonios Papanicolau-Sengos, Fatima Karzai, Mark R. Gilbert, Kenneth Aldape, Padma S. Rajagopal, Tuvik Beker, Eytan Ruppin, Ranit Aharonov

January, 2023

Precision oncology is gradually advancing into mainstream clinical practice, demonstrating significant survival benefits. However, eligibility and response rates remain limited in many cases, calling for better predictive biomarkers. Here we present ENLIGHT, a transcriptomics-based computational approach that identifies clinically relevant genetic interactions and uses them to predict a patient’s response to a variety of therapies in multiple cancer types, without training on previous treatment response data. We study ENLIGHT in two translationally oriented scenarios: Personalized Oncology (PO), aimed at prioritizing treatments for a single patient, and Clinical Trial Design (CTD), selecting the most likely responders in a patient cohort. Evaluating ENLIGHT’s performance on 21 blinded clinical trial datasets in the PO setting, we show that it can effectively predict a patient’s treatment response across multiple therapies and cancer types (obtaining an aggregate odds ratio of 2.59). Its prediction accuracy is better than previously published transcriptomics-based signatures and is comparable to that of supervised predictors developed for specific indications and drugs, without requiring specific response data for training. In combination with the IFN-γ signature, ENLIGHT achieves an odds ratio larger than 4 in predicting patients’ response to immune checkpoint therapy. In the CTD scenario, we demonstrate that ENLIGHT can potentially enhance clinical trial success for immunotherapies and other monoclonal antibodies (mAb) by excluding non-responders, while overall achieving more than 90% of the response rate attainable under an optimal exclusion strategy. In sum, in this retrospective study, ENLIGHT demonstrably enhances the ability to predict therapeutic response across multiple cancer types from the bulk tumor transcriptome.  

 

Read the full publication in Med    

 

Fibrolamellar carcinoma transcriptomic-based treatment prediction: complete response after nivolumab and ipilimumab

Raanan Berger, Gal Dinstag, Omer Tirosh, Eyal Schiff, David Kleiner, Kenneth D Aldape, Eytan Ruppin, Tuvik Beker and Razelle Kurzrock

December, 2022

Fibrolamellar carcinoma (FLC) is a rare cancer of the liver that most commonly affects children and young adults. There is no clear standard of care for the disease, whose response to treatment seems to be very different from that of hepatocellular carcinoma. We present a case of FLC in a patient in her mid 30s that recurred and persisted despite resection and multiple lines of treatment. Following transcriptomic analysis, a combination of ipilimumab (anti-CTLA4) and nivolumab (anti-PD-1) led to complete remission, although common biomarkers for immune checkpoint blockade were all negative in this case. The patient is still in remission. Here, combined checkpoint blockade guided by novel transcriptomic analysis led to complete remission after failure of several lines of treatment.  

 

Read the full publication in JITC    

 

Synthetic lethality-based prediction of cancer treatment response from histopathology images

Danh-Tai Hoang, Gal Dinstag, Leandro C. Hermida, Doreen S. Ben-Zvi, Efrat Elis, Katherine Caley, Sanju Sinha, Neelam Sinha, Christopher H. Dampier, Tuvik Beker, Kenneth Aldape, Ranit Aharonov, Eric A. Stone, Eytan Ruppin

June, 2022

Advances in artificial intelligence have paved the way for predicting cancer patients’ survival and response to treatment from hematoxylin and eosin (H&E)-stained tumor slides. Extant approaches do so either directly from the H&E images or via prediction of actionable mutations and gene fusions. Here we present the first genetic interactions (GI)-based approach for predicting patient response to treatment, founded on two conceptual steps: (1) First, we build DeepPT, a deep-learning framework that predicts tumor gene expression from H&E slides, and subsequently, (2) we apply ENLIGHT – a previously published GI based approach – to predict patient treatment response from the inferred tumor expression. DeepPT was trained on images and corresponding transcriptomics data of TCGA breast, kidney, lung, and brain tumor samples. Testing DeepPT transcriptomics prediction ability, we find that it generalizes well to predicting the expression of two breast and brain cancer unseen independent datasets. Studying samples from a recently published large multi-omics breast cancer clinical trial, we applied ENLIGHT to the expression predicted by DeepPT from the tumor slides. We find that it successfully predicts true responders with a clinically meaningful hazard ratio of about six. These results put forward a general framework for predicting patient response to a broad array of targeted and checkpoint therapies from the histological images. If corroborated further, the new approach could augment the feasibility of precision oncology in developing countries and in other situations where comprehensive molecular profiling is not available.  

 

Read the full publication in BioRxiv    

 

 

Synthetic lethality-mediated precision oncology via the tumor transcriptome

Joo Sang Lee, Nishanth Ulhas Nair, Gal Dinstag, Lesley Chapman, Youngmin Chung, Kun Wang, Sanju Sinha, Hongui Cha, Dasol Kim, Alexander V. Schperberg, Ajay Srinivasan Vladimir Lazar, Eitan Rubin, Sohyun Hwang, Raanan Berger, Tuvik Beker, Ze’ev Ronai, Sridhar Hannenhalli, Mark R. Gilbert, Razelle Kurzrock, Se-Hoon Lee, Kenneth Aldape Eytan Ruppin

April 29, 2021

Precision oncology has made significant advances, mainly by targeting actionable mutations in cancer driver genes. Aiming to expand treatment opportunities, recent studies have begun to explore the utility of tumor transcriptome to guide patient treatment. Here, we introduce SELECT (synthetic lethality and rescue-mediated precision oncology via the transcriptome), a precision oncology framework harnessing genetic interactions to predict patient response to cancer therapy from the tumor transcriptome. SELECT is tested on a broad collection of 35 published targeted and immunotherapy clinical trials from 10 different cancer types. It is predictive of patients’ response in 80% of these clinical trials and in the recent multi-arm WINTHER trial. The predictive signatures and the code are made publicly available for academic use, laying a basis for future prospective clinical studies.  

 

Read the full publication in Cell    

 

Synthetic lethality across normal tissues is strongly associated with cancer risk, onset, and tumor suppressor specificity

Kuoyuan Cheng, Nishanth Ulhas Nair, Joo Sang Lee, Eytan Ruppin

January 1, 2021

Various characteristics of cancers exhibit tissue specificity, including lifetime cancer risk, onset age, and cancer driver genes. Previously, the large variation in cancer risk across human tissues was found to strongly correlate with the number of stem cell divisions and abnormal DNA methylation levels. Here, we study the role of synthetic lethality in cancer risk. Analyzing normal tissue transcriptomics data in the Genotype-Tissue Expression project, we quantify the extent of co-inactivation of cancer synthetic lethal (cSL) gene pairs and find that normal tissues with more down-regulated cSL gene pairs have lower and delayed cancer risk. Consistently, more cSL gene pairs become up-regulated in cells treated by carcinogens and throughout premalignant stages in vivo. We also show that the tissue specificity of numerous tumor suppressor genes is associated with the expression of their cSL partner genes across normal tissues. Overall, our findings support the possible role of synthetic lethality in tumorigenesis.  

Read the full publication in Science Advances    

Synthetic lethal combination targeting BET uncovered intrinsic susceptibility of TNBC to ferroptosis

Nandini Verma, Yaron Vinik, Ashish Saroha, Nishanth Ulhas Nair, Eytan Ruppin, Gordon Mills, Thomas Karn, Vinay Dubey, Lohit Khera, Harsha Raj, Flavio Maina, Sima Lev

August 21, 2020

Identification of targeted therapies for TNBC is an urgent medical need. Using a drug combination screen reliant on synthetic lethal interactions, we identified clinically relevant combination therapies for different TNBC subtypes. Two drug combinations targeting the BET family were further explored. The first, targeting BET and CXCR2, is specific for mesenchymal TNBC and induces apoptosis, whereas the second, targeting BET and the proteasome, is effective for major TNBC subtypes and triggers ferroptosis. Ferroptosis was induced at low drug doses and was associated with increased cellular iron and decreased glutathione levels, concomitant with reduced levels of GPX4 and key glutathione biosynthesis genes. Further functional studies, analysis of clinical datasets, and breast cancer specimens revealed a unique vulnerability of TNBC to ferroptosis inducers, enrichment of ferroptosis gene signature, and differential expression of key proteins that increase labile iron and decrease glutathione levels. This study identified potent combination therapies for TNBC and unveiled ferroptosis as a promising therapeutic strategy.  

 

Read the full publication in Science Advances    

 

Translational reprogramming marks adaptation to asparagine restriction in cancer

Gaurav Pathria, Joo Sang Lee, Erez Hasnis, Kristofferson Tandoc, David A. Scott, Sachin Verma, Yongmei Feng, Lionel Larue, Avinash D. Sahu, Ivan Topisirovic, Eytan Ruppin & Ze’ev A. Ronai

November, 2019

While amino acid restriction remains an attractive strategy for cancer therapy, metabolic adaptations limit its effectiveness. Here we demonstrate the role of translational reprogramming in the survival of asparagine-restricted cancer cells. Asparagine limitation in melanoma and pancreatic cancer cells activates receptor tyrosine kinase–MAPK signalling as part of a feedforward mechanism involving mammalian target of rapamycin complex 1 (mTORC1)-dependent increase in MAPK-interacting kinase 1 (MNK1) and eukaryotic translation initiation factor 4E (eIF4E), resulting in enhanced translation of activating transcription factor 4 (ATF4) mRNA. MAPK inhibition attenuates translational induction of ATF4 and the expression of its target asparagine synthetase (ASNS), sensitizing melanoma and pancreatic tumours to asparagine restriction, reflected in the inhibition of their growth. Correspondingly, low ASNS expression is among the top predictors of response to inhibitors of MAPK signalling in patients with melanoma and is associated with favourable prognosis when combined with low MAPK signalling activity. These studies reveal an axis of adaptation to asparagine deprivation and present a rationale for the clinical evaluation of MAPK inhibitors in combination with asparagine restriction approaches.  

 

Read the full publication in Nature Cell Biology    

 

 

Beyond Synthetic Lethality: Charting the Landscape of Pairwise Gene Expression States Associated with Survival in Cancer

Assaf Magen, Avinash Das Sahu, Joo Sang Lee, Mahfuza Sharmin, Alexander Lugo, J. Silvio Gutkind, Alejandro A. Schäffer, Eytan Ruppin

July, 2019

The phenotypic effect of perturbing a gene’s activity depends on the activity level of other genes, reflecting the notion that phenotypes are emergent properties of a network of functionally interacting genes. In the context of cancer, contemporary investigations have primarily focused on just one type of functional relationship between two genes-synthetic lethality (SL). Here, we define the more general concept of “survival-associated pairwise gene expression states” (SPAGEs) as gene pairs whose joint expression levels are associated with survival. We describe a data-driven approach called SPAGE-finder that when applied to The Cancer Genome Atlas (TCGA) data identified 71,946 SPAGEs spanning 12 distinct types, only a minority of which are SLs. The detected SPAGEs explain cancer driver genes’ tissue specificity and differences in patients’ response to drugs and stratify breast cancer tumors into refined subtypes. These results expand the scope of cancer SPAGEs and lay a conceptual basis for future studies of SPAGEs and their translational applications.  

 

Read the full publication in Cell    

 

 

A Platform of Synthetic Lethal Gene Interaction Networks Reveals that the GNAQ Uveal Melanoma Oncogene Controls the Hippo Pathway through FAK

 

Xiaodong Feng, Nadia Arang, Damiano Cosimo Rigiracciolo, Joo Sang Lee, Huwate Yeerna, Zhiyong Wang, Simone Lubrano, Ayush Kishore, Jonathan A Pachter, Gabriele M König, Marcello Maggiolini, Evi Kostenis, David D Schlaepfer, Pablo Tamayo, Qianming Chen, Eytan Ruppin, J Silvio Gutkind

March, 2019

Activating mutations in GNAQ/GNA11, encoding Gαq G proteins, are initiating oncogenic events in uveal melanoma (UM). However, there are no effective therapies for UM. Using an integrated bioinformatics pipeline, we found that PTK2, encoding focal adhesion kinase (FAK), represents a candidate synthetic lethal gene with GNAQ activation. We show that Gαq activates FAK through TRIO-RhoA non-canonical Gαq-signaling, and genetic ablation or pharmacological inhibition of FAK inhibits UM growth. Analysis of the FAK-regulated transcriptome demonstrated that GNAQ stimulates YAP through FAK. Dissection of the underlying mechanism revealed that FAK regulates YAP by tyrosine phosphorylation of MOB1, inhibiting core Hippo signaling. Our findings establish FAK as a potential therapeutic target for UM and other Gαq-driven pathophysiologies that involve unrestrained YAP function.  

 

Read the full publication on Cell    

 

 

Genome-wide prediction of synthetic rescue mediators of resistance to targeted and immunotherapy

Avinash Das Sahu, Joo S Lee, Zhiyong Wang, Gao Zhang, Ramiro Iglesias-Bartolome, Tian Tian, Zhi Wei, Benchun Miao, Nishanth Ulhas Nair, Olga Ponomarova, Adam A Friedman, Arnaud Amzallag, Tabea Moll, Gyulnara Kasumova, Patricia Greninger, Regina K Egan, Leah J Damon, Dennie T Frederick, Livnat Jerby-Arnon, Allon Wagner, Kuoyuan Cheng, Seung Gu Park, Welles Robinson, Kevin Gardner, Genevieve Boland, Eytan Ruppin

March, 2019

Most patients with advanced cancer eventually acquire resistance to targeted therapies, spurring extensive efforts to identify molecular events mediating therapy resistance. Many of these events involve synthetic rescue (SR) interactions, where the reduction in cancer cell viability caused by targeted gene inactivation is rescued by an adaptive alteration of another gene (the rescuer). Here, we perform a genome-wide in silico prediction of SR rescuer genes by analyzing tumor transcriptomics and survival data of 10,000 TCGA cancer patients. Predicted SR interactions are validated in new experimental screens. We show that SR interactions can successfully predict cancer patients’ response and emerging resistance. Inhibiting predicted rescuer genes sensitizes resistant cancer cells to therapies synergistically, providing initial leads for developing combinatorial approaches to overcome resistance proactively. Finally, we show that the SR analysis of melanoma patients successfully identifies known mediators of resistance to immunotherapy and predicts novel rescuers.  

 

Read the full publication in Molecular Systems Biology    

 

Harnessing synthetic lethality to predict the response to cancer treatment

 

Joo Sang Lee, Avinash Das, Livnat Jerby-Arnon, Rand Arafeh, Noam Auslander, Matthew Davidson, Lynn McGarry, Daniel James, Arnaud Amzallag, Seung Gu Park, Kuoyuan Cheng, Welles Robinson, Dikla Atias, Chani Stossel, Ella Buzhor, Gidi Stein, Joshua J. Waterfall, Paul S. Meltzer, Talia Golan, Sridhar Hannenhalli, Eyal Gottlieb, Cyril H. Benes, Yardena Samuels, Emma Shanks & Eytan Ruppin

June, 2018

While synthetic lethality (SL) holds promise in developing effective cancer therapies, SL candidates found via experimental screens often have limited translational value. Here we present a data-driven approach, ISLE (identification of clinically relevant synthetic lethality), that mines TCGA cohort to identify the most likely clinically relevant SL interactions (cSLi) from a given candidate set of lab-screened SLi. We first validate ISLE via a benchmark of large-scale drug response screens and by predicting drug efficacy in mouse xenograft models. We then experimentally test a select set of predicted cSLi via new screening experiments, validating their predicted context-specific sensitivity in hypoxic vs normoxic conditions and demonstrating cSLi’s utility in predicting synergistic drug combinations. We show that cSLi can successfully predict patients’ drug treatment response and provide patient stratification signatures. ISLE thus complements existing actionable mutation-based methods for precision cancer therapy, offering an opportunity to expand its scope to the whole genome.

 

Read the full publication in Nature Communications    

 

 

Predicting Cancer-Specific Vulnerability via Data-Driven Detection of Synthetic Lethality

Livnat Jerby-Arnon, Nadja Pfetzer, Yedael Y. Waldman, Lynn McGarry, Daniel James, Emma Shanks,  Brinton Seashore-Ludlow, Adam Weinstock, Tamar Geiger, Paul A. Clemons, Eyal Gottlieb, Eytan Ruppin.

August , 2014

Synthetic lethality (SL), an important type of genetic interaction, can provide useful insight into the target identification process for the development of anticancer therapeutics. Although several well-established SL gene pairs have been verified to be conserved in humans, most SL interactions remain cell-line specific. Here, we demonstrated that the cell-line-specific gene expression profiles derived from the shRNA perturbation experiments performed in the LINCS L1000 project can provide useful features for predicting SL interactions in humans. In this paper, we developed a semi-supervised neural network-based method called EXP2SL to accurately identify SL interactions from the L1000 gene expression profiles. Through a systematic evaluation of the SL datasets of three different cell lines, we demonstrated that our model achieved better performance than the baseline methods and verified the effectiveness of using the L1000 gene expression features and the semi-supervised training technique in SL prediction.

 

Read the full publication in Cell