Biology subjects

Haibe-Kains, B.

Publications and source records attributed to Haibe-Kains, B..

At least 19 recordsLinked to original sources

bioRxiv2025

Distributed Biomarker Discovery for Immuno-Oncology

BackgroundIdentifying robust predictive biomarkers for immuno-oncology (IO) therapy response remains challenging due to the complexity of tumor host interactions and the limited availability of public datasets. While achieving clinical relevance requires biomarkers that generalize across diverse datasets, privacy concerns surrounding sensitive clinical and molecular data further restrict data sharing, impeding progress in biomarker discovery. Materials and MethodsWe developed a scalable, cloud-based distributed pipeline, implemented as an in-silico simulation of federated learning, to enable privacy-preserving integration of clinical and molecular data across institutions. This approach supports the development of genomic prediction algorithms without sharing sensitive patient data. Using curated gene expression profiles, we evaluated the association of tumor microenvironment and IO gene expression signatures with clinical outcomes across pan-cancer, cancer-specific, and treatment-specific settings, and developed a multivariable distributed model to predict IO response. ResultsWe implemented an integrated meta-analysis pipeline using harmonized data across 18 datasets comprising 967 patients and seven cancer types treated with PD-1/PD-L1, CTLA-4, or combination immunotherapies. This approach identified biomarkers specific to lung cancer and CTLA-4 therapy previously unreported in pan-cancer and pan-IO analyses, highlighting the value of context-specific associations. Similar gene expression signatures were associated with progression-free survival and response in both pan-cancer and melanoma datasets, with notable overlap between dual checkpoint blockade and PD-1/PD-L1 therapies. Distributed multivariable XGBoost models outperformed the average of locally trained models, indicating improved generalizability. ConclusionWe present a new distributed pipeline for biomarker discovery in immuno-oncology that preserves patient privacy through federated analysis and secure data handling. This study highlights the importance of larger, diverse datasets to refine cancer- and treatment-specific biomarkers, paving the way for more precise and personalized IO therapies.

bioinformatics↗

bioRxiv2025

Functional landscape of circular RNAs in human cancer cells

Circular RNAs (circRNAs) constitute a novel class of noncoding RNAs showcasing distinct tissue- and cell-specific expression patterns. Despite the extensive profiling of circRNAs, their individual functions remain poorly understood. To fill this gap, we designed a genome-wide library of 65,300 shRNAs, targeting 9,663 clinically relevant circRNAs and 3,981 of their linear parental genes, and conducted functional screening in seven types of human cancer. We identified a total of 1,342 essential circRNAs (13.9% screened) that impact cell proliferation in at least one cell line, and in 96.5% of the cases, the linear counterparts are not essential. While a shared common subset emerges as functional regulators across all examined cell lines, the majority of circRNAs are functional in a cell type-specific manner. For a comprehensive presentation of the functional circRNA landscape in cancer, we introduce FunCirc, an online database encompassing functional circRNAs across cancer cell lines, coupled with circRNA expression profiles from diverse cancer and tissue types. Our work enhances the understanding of circRNA functions in cancer and provides the scientific community with a resource to further investigate their intricate roles.

cancer biology↗

bioRxiv2025

Circular RNA profiling and functional screening in breast cancer identify circNSD1 as a suppressor of tumor autophagy

Circular RNAs (circRNAs) have emerged as critical regulators of cell biology. However, their function in breast cancer remains elusive. Herein, through circRNA profiling of 38 breast tumors and 10 benign breast tissues, we identified 509 of differentially expressed circRNAs. Integration with transcriptome-wide functional screening of approximately 10,000 circRNAs pinpointed circNSD1(6) as a top ranked tumor suppressor in breast cancer. circNSD1(6) is downregulated in tumors and suppression of the circular form, but not the linear form, promotes breast cancer proliferation and tumor growth. Mechanistically, circNSD1(6) interacts with autophagy receptor SQSTM1/p62 and inhibits the oligomerization of p62, thereby suppressing p62 body formation and p62-dependent autophagy. Furthermore, circNSD1(6) ameliorates p62-mediated Keap1 sequestration, hence suppresses the Nrf2 pathway and oxidative stress tolerance. Our study provides a landscape view of the transcription and function of circRNA in breast cancer, and uncovers the crucial role of circNSD1(6) in autophagy and antioxidant stress.

cancer biology↗

bioRxiv2024

MorphoDiff: Cellular Morphology Painting with Diffusion Models

Understanding cellular responses to external stimuli is critical for parsing biological mechanisms and advancing therapeutic development. High-content image-based assays provide a cost-effective approach to examine cellular phenotypes induced by diverse interventions, which offers valuable insights into biological processes and cellular states. In this paper, we introduce MorphoDiff, a generative pipeline to predict high-resolution cell morphological responses under different conditions based on perturbation encoding. To the best of our knowledge, MorphoDiff is the first framework capable of producing guided, high-resolution predictions of cell morphology that generalize across both chemical and genetic interventions. The model integrates perturbation embeddings as guiding signals within a 2D latent diffusion model. The comprehensive computational, biological, and visual validations across three open-source Cell Painting datasets show that MorphoDiff can generate high-fidelity images and produce meaningful biology signals under various interventions. We envision the model will facilitate efficient in silico exploration of perturbational landscapes towards more effective drug discovery studies.

bioinformatics↗

bioRxiv2023

Reusability Report: Few-shot learning creates predictive models of drug response that translate from high-throughput screens to individual patients

Machine learning (ML) and artificial intelligence (AI) methods are increasingly used in personalized medicine, including precision oncology. Ma et al. (Nature Cancer 2021) developed a new method c alled "Transfer of Cell Line Response Prediction" (TCRP) to train predictors of drug response in cancer cell lines and optimize their performance in higher complex cancer model systems via few-shot learning. TCRP was presented as a successful modeling approach in multiple case studies. Given the importance of this approach to assist clinicians in their treatment decision process, we sought to reproduce independently the authors findings and improve the reusability of TCRP in new case studies, including validation in clinical trial datasets, a high bar for drug response prediction. Our results support the superiority of TCRP over established statistical and machine learning approaches in preclinical and clinical settings. We developed new resources to increase the reusability of the TCRP model for future improvements and validation studies.

bioinformatics↗

bioRxiv2023

Similarity metric learning on perturbational datasets improves functional identification of perturbations

Analysis of high-throughput perturbational datasets, including the Next Generation Connectivity Map (L1000) and the Cell Painting projects, uses similarity metrics to identify perturbations or disease states that induce similar changes in the biological feature space. Similarities among perturbations are then used to identify drug mechanisms of action, to nominate therapeutics for a particular disease, and to construct bio-logical networks among perturbations and genes. Standard similarity metrics include correlations, cosine distance and gene set enrichment methods, but these methods operate on the measured features without refinement by transforming the measurement space. We introduce Perturbational Metric Learning (PeML), a weakly supervised similarity metric learning method to learn a data-driven similarity function that maximizes discrimination of replicate signatures by transforming the biological measurements into an intrinsic, dataset-specific basis. The learned similarity functions show substantial improvement for recovering known biological relationships, like mechanism of action identification. In addition to capturing a more meaningful notion of similarity, data in the transformed basis can be used for other analysis tasks, such as classification and clustering. Similarity metric learning is a powerful tool for the analysis of large biological datasets.

bioinformatics↗

bioRxiv2023

GraphComm: A Graph-based Deep Learning Method to Predict Cell-Cell Communication in single-cell RNAseq data

Interactions between cells coordinate various functions across cell-types in health and disease states. Novel single-cell techniques enable deep investigation of cellular crosstalk at single-cell resolution. Cell-cell communication (CCC) is mediated by underlying gene-gene networks, however most current methods are unable to account for complex interactions within the cell as well as incorporate the effect of pathway and protein complexes on interactions. This results in the inability to infer overarching signalling patterns within a dataset as well as limit the ability to successfully explore other data types such as spatial cell dimension. Therefore, to represent transcriptomic data as intricate networks, complementing gene expression with information from cells to ligands and receptors for relevant cell-cell communication inference, we present GraphComm - a new graph-based deep learning method for predicting cell-cell communication in single-cell RNAseq datasets. GraphComm improves CCC inference by capturing detailed information such as cell location and intracellular signalling patterns from a database of more than 30,000 protein interaction pairs. With this framework, GraphComm is able to predict biologically relevant results in datasets previously validated for CCC, datasets that have undergone chemical or genetic perturbations and datasets with spatial cell information.

cell biology↗

bioRxiv2023

GENIUS: GEnome traNsformatIon and spatial representation of mUltiomicS data

The application of next-generation sequencing (NGS) has transformed cancer research. As costs have decreased, NGS has increasingly been applied to generate multiple layers of molecular data from the same samples, covering genomics, transcriptomics, and methylomics. Integrating these types of multi-omics data in a combined analysis is now becoming a common issue with no obvious solution, often handled on an ad-hoc basis, with multi-omics data arriving in a tabular format and analyzed using computationally intensive statistical methods. These methods particularly ignore the spatial orientation of the genome and often apply stringent p-value corrections that likely result in the loss of true positive associations. Here, we present GENIUS (GEnome traNsformatIon and spatial representation of mUltiomicS data), a framework for integrating multi-omics data using deep learning models developed for advanced image analysis. The GENIUS framework is able to transform multi-omics data into images with genes displayed as spatially connected pixels and successfully extract relevant information with respect to the desired output. Here, we demonstrate the utility of GENIUS by applying the framework to multi-omics datasets from the Cancer Genome Atlas. Our results are focused on predicting the development of metastatic cancer from primary tumors, and demonstrate how through model inference, we are able to extract the genes which are driving the model prediction and likely associated with metastatic disease progression. We anticipate our framework to be a starting point and strong proof of concept for multi-omics data transformation and analysis without the need for statistical correction.

bioinformatics↗

bioRxiv2023

Detection of circular RNAs and their potential as biomarkers predictive of drug response

The introduction of high-throughput sequencing technologies has allowed for comprehensive RNA species detection, both coding and non-coding, which opened new avenues for the discovery of predictive and prognostic biomarkers. However the consistency of the detection of different RNA species depends on the RNA selection protocol used for RNA-sequencing. While preliminary reports indicated that non-coding RNAs, in particular circular RNAs, constitute a rich source of biomarkers predictive of drug response, the reproducibility of this novel class of biomarkers has not been rigorously investigated. To address this issue, we assessed the inter- lab consistency of circular RNA expression in cell lines profiled in large pharmacogenomic datasets. We found that circular RNA expression quantified from rRNA-depleted RNA-seq data is stable and yields robust prognostic markers in cancer. On the other hand, quantification of the expression of circular RNA from poly(A)-selected RNA-seq data yields highly inconsistent results, calling into question results from previous studies reporting their potential as predictive biomarkers in cancer. We have also identified median expression of transcripts and transcript length as potential factors influencing the consistency of RNA detection. Our study provides a framework to quantitatively assess the stability of coding and non-coding RNA expression through the analysis of biological replicates within and across independent studies.

bioinformatics↗

bioRxiv2022

Meta-analysis of preclinical pharmacogenomic studies to discover robust and translatable biomarkers of drug response

Preclinical pharmacogenomic studies provide an opportunity to discover novel biomarkers for drug response. However, pharamcogenomic studies linking gene expression profiles to drug response do not always agree on the significance or strength of biomarkers. We apply a statistical meta-analysis approach to 7 large independent pharmacogenomic studies, testing for tissue-specific gene expression markers predictive of response among cancer cell lines. We found 4,338 statistically-significant biomarkers across 8 tissue types and 34 drugs. Significant biomarkers were found to be closer than random to drug targets in a gene network built on pathway co-membership (average distance of 2 vs 2.9). However, functional relationships with the drug target did not predict reproducibility across studies. To validate these biomarkers, we utilized 10 clinical datasets, allowing 42/4338 biomarkers to be assessed for clinical translation. Of the 42 candidate biomarkers, the expression of ODC1 was found to be significantly predictive of Paclitaxel response as a neoadjuvant treatment of breast carcinoma across 2 independent clinical studies of >200 patients each. We expect that as more clinical transcriptomics data matched with response are available, our results can be used to prioritize which genes to evaluate as clinical biomarkers of drug response.

cancer biology↗

bioRxiv2022

Similarity bias from averaging signatures from the Connectivity Map

In recent years, high-throughput perturbational datasets have become an important tool for rapidly characterizing the function of large collections of chemical compounds. To overcome the biological and technical noise in these experiments, researchers have used consensus signatures - averages of multiple experiments - to summarize the effects of perturbations. In this work, we demonstrate that consensus signatures on the L1000 Connectivity Map show a pervasive similarity bias: as more signatures are averaged, the resulting consensus signatures are increasingly similar to each other, regardless of whether the signatures are related. We show that the distribution of Pearsons correlation changes as a function of the number of signatures averaged. The artifactual similarity bias is caused by skewness in the data and a consequence of using median normalization on non-normal distributions. Furthermore, we show that mean normalization can partly remedy this similarity bias and improve power to identify associations. The similarity bias introduced by consensus signatures is an important potential confounder of analysis of perturbational datasets, and our practical solution could easily be applied by practitioners in the field to improve the analysis of the L1000 Connectivity Map.

bioinformatics↗

bioRxiv2021

PharmacoDB 2.0 : Improving scalability and transparency of in vitro pharmacogenomics analysis

Cancer pharmacogenomics studies provide valuable insights into disease progression and associations between genomic features and drug response. PharmacoDB integrates multiple cancer pharmacogenomics datasets profiling approved and investigational drugs across cell lines from diverse tissue types. The web-application enables users to efficiently navigate across datasets, view and compare drug dose-response data for a specific drug-cell line pair. In the new version of PharmacoDB (version 2.0, https://pharmacodb.ca/), we present: (i) new datasets such as NCI-60, the Profiling Relative Inhibition Simultaneously in Mixtures (PRISM) dataset, as well as updated data from the Genomics of Drug Sensitivity in Cancer (GDSC) and the Genentech Cell Line Screening Initiative (gCSI); (ii) implementation of FAIR data pipelines using ORCESTRA and PharmacoDI; (iii) enhancements to drug response analysis such as tissue distribution of dose-response metrics and biomarker analysis; (iv) improved connectivity to drug and cell line databases in the community. The web interface has been rewritten using a modern technology stack to ensure scalability and standardization to accommodate growing pharmacogenomics datasets. PharmacoDB 2.0 is a valuable tool for mining pharmacogenomics datasets, comparing and assessing drug response phenotypes of cancer models. HIGHLIGHTSO_LIPharmacoDB 2.0 includes new and updated large pharmacogenomic datasets. The data processing for PharmacoDB is made fully reproducible through the use of the ORCESTRA platform and automated data ingestion pipelines C_LIO_LIThe new release contains enriched annotations for drugs and cell lines via connectivity to external databases, as well as new analytical methods for tissue-specific and pan-cancer biomarker discovery C_LIO_LIThe new version of PharmacoDB incorporates a scalable and reproducible framework that can accelerate the implementation of analytical pipelines including machine learning/AI for biomarker discovery in the future C_LI

bioinformatics↗

bioRxiv2021

Drug Sensitivity Prediction From Cell Line-Based Pharmacogenomics Data: Guidelines for Developing Machine Learning Models

The goal of precision oncology is to tailor treatment for patients individually using the genomic profile of their tumors. Pharmacogenomics datasets such as cancer cell lines are among the most valuable resources for drug sensitivity prediction, a crucial task of precision oncology. Machine learning methods have been employed to predict drug sensitivity based on the multiple omics data available for large panels of cancer cell lines. However, there are no comprehensive guidelines on how to properly train and validate such machine learning models for drug sensitivity prediction. In this paper, we introduce a set of guidelines for different aspects of training gene expression-based predictors using cell line datasets. These guidelines provide extensive analysis of the generalization of drug sensitivity predictors, and challenge many current practices in the community including the choice of training dataset and measure of drug sensitivity. Application of these guidelines in future studies will enable the development of more robust preclinical biomarkers.

bioinformatics↗

bioRxiv2020

ORCESTRA: a platform for orchestrating and sharing high-throughput pharmacogenomic analyses

Reproducibility is essential to open science, as there is limited relevance for findings that can not be reproduced by independent research groups, regardless of its validity. It is therefore crucial for scientists to describe their experiments in sufficient detail so they can be reproduced, scrutinized, challenged, and built upon. However, the intrinsic complexity and continuous growth of biomedical data makes it increasingly difficult to process, analyze, and share with the community in a FAIR (findable, accessible, interoperable, and reusable) manner. To overcome these issues, we created a cloud-based platform called ORCESTRA (orcestra.ca), which provides a flexible framework for the reproducible processing of multimodal biomedical data. It enables processing of clinical, genomic and perturbation profiles of cancer samples through automated processing pipelines that are user-customizable. ORCESTRA creates integrated and fully documented data objects with persistent identifiers (DOI) and manages multiple dataset versions, which can be shared for future studies.

bioinformatics↗

bioRxiv2020

Computational pharmacogenomics screen identifies synergistic statin-compound combinations as anti-breast cancer therapies

Statins are a family of FDA-approved cholesterol-lowering drugs that inhibit the rate-limiting enzyme of the metabolic mevalonate pathway, which have been shown to have anti-cancer activity. As therapeutic efficacy is increased when drugs are used in combination, we sought to identify agents, like dipyridamole, that potentiate statin-induced tumor cell death. As an antiplatelet agent dipyridamole will not be suitable for all cancer patients. Thus, we developed an integrative pharmacogenomics pipeline to identify agents that were similar to dipyridamole at the level of drug structure, in vitro sensitivity and molecular perturbation. To enrich for compounds expected to target the mevalonate pathway, we took a pathway-centric approach towards computational selection, which we called mevalonate drug network fusion (MVA-DNF). We validated two of the top ranked compounds, nelfinavir and honokiol and demonstrated that, like dipyridamole, they synergize with fluvastatin to potentiate tumor cell death by blocking the restorative feedback loop. This is achieved by inhibiting activation of the key transcription factor that induces mevalonate pathway gene transcription, sterol regulatory element-binding protein 2 (SREBP2). Mechanistically, the synergistic response of fluvastatin+nelfinavir and fluvastatin+honokiol was associated with similar transcriptomic and proteomic pathways, indicating a similar mechanism of action between nelfinavir and honokiol when combined with fluvastatin. Further analysis identified the canonical epithelial-mesenchymal transition (EMT) gene, E-cadherin as a biomarker of these synergistic responses across a large panel of breast cancer cell lines. Thus, our computational pharmacogenomic approach can identify novel compounds that phenocopy a compound of interest in a pathway-specific manner. Significance StatementWe provide a rapid and cost-effective strategy to expand a class of drugs with a similar phenotype. Our parent compound, dipyridamole, potentiated statin-induced tumor cell death by blocking the statin-triggered restorative feedback response that dampens statins pro-apoptotic activity. To identify compounds with this activity we performed a pharmacogenomic analysis to distinguish agents similar to dipyridamole in terms of structure, cell sensitivity and molecular perturbations. As dipyridamole has many reported activities, we focused our molecular perturbation analysis on the pathway inhibited by statins, the metabolic mevalonate pathway. Our strategy was successful as we validated nelfinavir and honokiol as dipyridamole-like drugs at both the phenotypic and molecular levels. Our pathway-specific pharmacogenomics approach will have broad applicability.

cancer biology↗

bioRxiv2020

KuLGaP: A Selective Measure for Assessing Therapy Response in Patient-Derived Xenografts

Quantifying response to drug treatment in mouse models of human cancer is important for treatment development and assignment, and yet remains a challenging task. A preferred measure to quantify this response should take into account as much of the experimental data as possible, i.e. both tumor size over time and the variation among replicates. We propose a theoretically grounded measure, KuLGaP, to compute the difference between the treatment and control arms. KuLGaP is more selective than currently existing measures, reduces the risk of false positive calls and improves translation of the lab results to clinical practice.

bioinformatics↗

bioRxiv2020

Bimodality of gene expression in cancer patient tumors as interpretable biomarkers for drug sensitivity

Identifying biomarkers predictive of cancer cells response to drug treatment constitutes one of the main challenges in precision oncology. Recent large-scale cancer pharmacogenomic studies have boosted the research for finding predictive biomarkers by profiling thousands of human cancer cell lines at the molecular level and screening them with hundreds of approved drugs and experimental chemical compounds. Many studies have leveraged these data to build predictive models of response using various statistical and machine learning methods. However, a common challenge in these methods is the lack of interpretability as to how they make the predictions and which features were the most associated with response, hindering the clinical translation of these models. To alleviate this issue, we develop a new machine learning pipeline based on the recent LOBICO approach that explores the space of bimodally expressed genes in multiple large in vitro pharmacogenomic studies and builds multivariate, nonlinear, yet interpretable logic-based models predictive of drug response. Using our method, we used a compendium of three of the largest pharmacogenomic data sets to build robust and interpretable models for 101 drugs that span 17 drug classes with high validation rate in independent datasets.

cancer biology↗

bioRxiv2019

Meta-gene markers predict meningioma recurrence with high accuracy

BackgroundMeningiomas, the most common adult brain tumors, recur in up to half of cases[1-3]. This requires timely intervention and therefore accurate risk assessment of recurrence is essential. Our current practice relies heavily on histological grade and extent of surgical excision to predict meningioma recurrence. However, prediction accuracy can be as poor as 50% for low or intermediate grade tumors, which constitute more than 90% of all cases. Moreover, attempts to find molecular markers, particularly for predicting recurrence of intermediate tumors have been impeded by low[4] or heterogenous[5-8] genetic signal. We therefore sought to apply systems-biology approaches to transcriptomic data to better predict meningioma recurrence.\n\nMethodsWe apply gene co-expression networks to a cohort of 252 adult patients from the publicly-available genetic repository Gene Expression Omnibus. Resultant gene clusters (\"modules\") were represented by the first principle component of their expression levels, and their ability to predict recurrence assessed with a logistic regression model. External validation was done using an independent merged cohort of 108 patients using the same modules. We used the bioinformatics database DAVID[9] to examine the gene ontology associations of each module.\n\nResultsUsing gene co-expression analysis, we were able predict tumor recurrence with high accuracy using a single module which mapped to cell cycle-related functions (AUC of 0.81 {+/-} 0.09 in external validation). Further, the module maintained its accuracy with removal of up to 20 of its most individually predictive genes, suggesting robustness against noise at the single gene level.\n\nConclusionsWith the easy accessibility of gene panels in healthcare diagnostics, our results offer a basis for routine molecular testing in meningioma management.

bioinformatics↗