A5009196285- Bioinformatics and Genomic Networks
- Computational Drug Discovery Methods
- Gene expression and cancer classification
- SARS-CoV-2 detection and testing
- COVID-19 Clinical Research Studies
- Protein Structure and Dynamics
- Advanced biosensing and bioanalysis techniques
- Cell Image Analysis Techniques
- Single-cell and spatial transcriptomics
- Machine Learning in Materials Science
- Genomics and Chromatin Dynamics
- Gene Regulatory Network Analysis
- Wireless Communication Networks Research
- interferon and immune responses
- Neutrophil, Myeloperoxidase and Oxidative Mechanisms
- Machine Learning in Bioinformatics
- Microbial Metabolic Engineering and Bioproduction
- Advanced Wireless Communication Techniques
- Machine Learning in Healthcare
- Chronic Obstructive Pulmonary Disease (COPD) Research
- Viral Infections and Immunology Research
- Immunodeficiency and Autoimmune Disorders
- Metabolomics and Mass Spectrometry Studies
- Inflammasome and immune disorders
- Long-Term Effects of COVID-19
McGill University
2018-2024
Mila - Quebec Artificial Intelligence Institute
2021-2024
McGill University Health Centre
2023-2024
Cancer Institute (WIA)
2024
Goodman (Japan)
2022-2024
University of Illinois Urbana-Champaign
2011-2020
University of Alberta
2009
We introduce GRouNdGAN, a gene regulatory network (GRN)-guided reference-based causal implicit generative model for simulating single-cell RNA-seq data, in silico perturbation experiments, and benchmarking GRN inference methods. Through the imposition of user-defined its architecture, GRouNdGAN simulates steady-state transient-state datasets where genes are causally expressed under control their regulating transcription factors (TFs). Training on six experimental reference datasets, we show...
We present Knowledge Engine for Genomics (KnowEnG), a free-to-use computational system analysis of genomics data sets, designed to accelerate biomedical discovery. It includes tools popular bioinformatics tasks such as gene prioritization, sample clustering, set analysis, and expression signature analysis. The specializes in "knowledge-guided" mining machine learning algorithms, which user-provided are analyzed light prior information about genes, aggregated from numerous knowledge bases...
Abstract Motivation Combination therapies have emerged as a treatment strategy for cancers to reduce the probability of drug resistance and improve outcomes. Large databases curating results many screening studies on preclinical cancer cell lines been developed, capturing synergistic antagonistic effects combination drugs in different lines. However, due high cost experiments sheer size possible combinations, these are quite sparse. This necessitates development transductive computational...
Pre-existing respiratory diseases may influence coronavirus disease (COVID-19) susceptibility and severity. However, the molecular mechanisms underlying airway epithelial response to severe acute syndrome 2 (SARS-CoV-2) infection severity in patients with chronic remain unelucidated. Using an vitro model of differentiated primary bronchial cells, we aimed investigate SARS-CoV-2 pre-existing cystic fibrosis (CF) obstructive pulmonary (COPD). Our study revealed reduced CF COPD epithelia...
Sequence-to-expression models have gained popularity in the past decade, enabling prediction of gene expression from genomic sequence alone. However, these typically do not take into account chromatin accessibility, a major factor limiting regulation. We hypothesized that supplying accessibility as an input feature would allow sequence-to-expression model to focus on important open regions genome. Using single-nucleus multiome RNA- and ATAC-sequencing data, we found predictive performance...
Prediction of clinical drug response (CDR) cancer patients, based on their and molecular profiles obtained prior to administration the drug, can play a significant role in individualized medicine. Machine learning models have potential address this issue but training them requires data from large number patients treated with each limiting feasibility. While databases preclinical in-vitro cell lines (CCLs) exist for many drugs, it is unclear whether samples be used predict CDR real patients....
Computational methods for the prediction of protein-protein interactions (PPIs), while important tools researchers, are plagued by challenges in generalizing to unseen proteins. Datasets used modelling predictions particularly predisposed information leakage and sampling biases.In this study, we introduce RAPPPID, a method Regularized Automatic Prediction Protein-Protein Interactions using Deep Learning. RAPPPID is twin Averaged Weight-Dropped Long Short-Term memory network which employs...
Abstract Motivation Interpretable deep learning (DL) models that can provide biological insights, in addition to accurate predictions, are of great interest the biomedical community. Recently, interpretable DL incorporate signaling pathways have been proposed for drug response prediction (DRP). While these improve interpretability, it is unclear whether this comes at cost less DRPs, or a improvement also be obtained. Results We comprehensively and systematically assessed four...
Abnormal coagulation and an increased risk of thrombosis are features severe COVID-19, with parallels proposed hemophagocytic lymphohistiocytosis (HLH), a life-threating condition associated hyperinflammation. The presence HLH was described in severely ill patients during the H1N1 influenza epidemic, presenting pulmonary vascular thrombosis. We tested hypothesis that genes causing primary regulate pathways linking thromboembolism to SARS-CoV-2 using novel network-informed computational...
The increasing number of publicly available databases containing drugs' chemical structures, their response in cell lines, and molecular profiles the lines has garnered attention to problem drug prediction. However, many existing methods do not fully leverage information that is shared among drugs with similar structure. As such, similarities terms line responses structures could prove be useful forming representations improve prediction accuracy.We present two deep learning approaches,...
Abstract Prediction of the response cancer patients to different treatments and identification biomarkers drug are two major goals individualized medicine. Here, we developed a deep learning framework called TINDL, completely trained on preclinical cell lines (CCLs), predict treatments. TINDL utilizes tissue-informed normalization account for tissue type tumors reduce statistical discrepancies between CCLs patient tumors. Moreover, by making black box interpretable, this model identifies...
Background Human glomerulonephritis (GN)—membranous nephropathy (MN), focal segmental glomerulosclerosis (FSGS) and IgA (IgAN), as well diabetic (DN) are leading causes of chronic kidney disease. In these glomerulopathies, distinct stimuli disrupt metabolic pathways in glomerular cells. Other pathways, including the endoplasmic reticulum (ER) unfolded protein response (UPR) autophagy, activated parallel to attenuate cell injury or promote repair. Methods We used publicly available datasets...
An overwhelming majority of protein-protein interaction (PPI) studies are conducted in a select few model organisms largely due to constraints time and cost the associated 'wet lab' experiments. In silico PPI inference methods ideal tools overcome these limitations, but often struggle with cross-species predictions. We present INTREPPPID, method that incorporates orthology data using new 'quintuplet' neural network, which is constructed five parallel encoders shared parameters. INTREPPPID...
Severe COVID-19 is associated with neutrophilic inflammation and immunothrombosis. Several members of the IL-17 cytokine family have been activation endothelium. Therefore, we investigated whether these cytokines were COVID-19.
Reconstruction of transcriptional regulatory networks (TRNs) is a powerful approach to unravel the gene expression programs involved in healthy and disease states cell. However, these are usually reconstructed independent phenotypic (or clinical) properties samples. Therefore, they may confound mechanisms that specifically related property with more general underlying full complement analyzed In this study, we develop method called InPheRNo identify "phenotype-relevant" TRNs. This based on...
Innate responses provide the first line of defense against viral infections, including influenza virus at mucosal surfaces. Communication and interaction between different host cells early stage infections determine quality magnitude immune invading virus. The release membrane-encapsulated extracellular vesicles (EVs), from cells, is defined as a refined system cell-to-cell communication. EVs contain diverse array biomolecules, microRNAs (miRNAs). We hypothesized that activation tracheal...
Identification of transcriptional regulatory mechanisms and signaling networks involved in the response host cells to infection by SARS-CoV-2 is a powerful approach that provides systems biology view gene expression programs COVID-19 may enable identification novel therapeutic targets strategies mitigate impact this disease. In study, our goal was identify network associated with changes between samples infected those are other respiratory viruses narrow results on enriched or specific...
Abstract We introduce GRouNdGAN, a gene regulatory network (GRN)-guided causal implicit generative model for simulating single-cell RNA-seq data, in-silico perturbation experiments, and benchmarking GRN inference methods. Through the imposition of user-defined in its architecture, GRouNdGAN simulates steady-state transient-state datasets where genes are causally expressed under control their regulating transcription factors (TFs). Training on three experimental datasets, we show that our...
ABSTRACT Background Prediction of the response cancer patients to different treatments and identification biomarkers drug sensitivity are two major goals individualized medicine. In this study, we developed a deep learning framework called TINDL, completely trained on preclinical cell lines, predict treatments. TINDL utilizes tissue-informed normalization account for tissue type tumours reduce statistical discrepancies between lines patient tumours. addition, model identifies small set genes...
We consider the problem of noiseless and noisy low- rank tensor completion from a set random linear measurements. In our derivations, we assume that entries belong to finite field arbitrary size reconstruction is based on minimization framework. The derived results show smallest number measurements needed for exact upper bounded by product rank, order, dimension cubic tensor. Furthermore, this condition also sufficient unique minimization. Similar bounds hold scenario, except scaling...
Abstract An overwhelming majority of protein-protein interaction (PPI) studies are conducted in a select few model organisms largely due to constraints time and cost the associated “wet lab” experiments. In silico PPI inference methods ideal tools overcome these limitations, but often struggle with cross-species predictions. We present INTREPPPID, method which incorporates orthology data using new “quintuplet” neural network, is constructed five parallel encoders shared parameters....
Abstract Linking DNA sequence to genomic function remains one of the grand challenges in genetics and genomics. Here, we combine large-scale single-molecule transcriptome sequencing diverse cancer cell lines with cutting-edge machine learning build LoRNA SH , an RNA foundation model that learns how nucleotide unspliced pre-mRNA dictates architecture—the relative abundances molecular structures mRNA isoforms. Owing its use StripedHyena architecture, handles extremely long inputs at base-pair...