A5084304645- Computational Drug Discovery Methods
- Single-cell and spatial transcriptomics
- Recommender Systems and Techniques
- Pharmacogenetics and Drug Metabolism
- Machine Learning in Bioinformatics
- Bioinformatics and Genomic Networks
- Innovative Microfluidic and Catalytic Techniques Innovation
- Biosimilars and Bioanalytical Methods
- Gene expression and cancer classification
- Cell Image Analysis Techniques
- SARS-CoV-2 and COVID-19 Research
- Human Mobility and Location-Based Analysis
- Hepatitis C virus research
- Advanced Graph Neural Networks
- Protein Structure and Dynamics
- SARS-CoV-2 detection and testing
- Complex Network Analysis Techniques
- Cancer Genomics and Diagnostics
- Music and Audio Processing
- Sparse and Compressive Sensing Techniques
- Diabetes and associated disorders
- Extracellular vesicles in disease
- Genomics and Chromatin Dynamics
- Advanced Computing and Algorithms
- Gene Regulatory Network Analysis
University of Pennsylvania
2022-2024
Indraprastha Institute of Information Technology Delhi
2018-2022
Indian Institute of Technology Delhi
2018-2022
Genome Institute of Singapore
2020-2022
Agency for Science, Technology and Research
2020-2022
Cancer Research Institute
2022
Abstract The emergence of single-cell RNA sequencing (scRNA-seq) technologies has enabled us to measure the expression levels thousands genes at resolution. However, insufficient quantities starting in individual cells cause significant dropout events, introducing a large number zero counts matrix. To circumvent this, we developed an autoencoder-based sparse gene matrix imputation method. AutoImpute, which learns inherent distribution input scRNA-seq data and imputes missing values...
Motivation: Single cell RNA sequencing has been proved to be revolutionary for its potential of zooming into complex biological systems. Genome wide expression analysis at single resolution, provides a window dynamics cellular phenotypes. This facilitates characterization transcriptional heterogeneity in normal and diseased tissues under various conditions. It also sheds light on development or emergence specific populations However, owing the paucity input RNA, typical data features high...
The identification of potential interactions between drugs and target proteins is crucial in pharmaceutical sciences. experimental validation genomic drug discovery laborious expensive; hence, there a need for efficient accurate in-silico techniques which can predict drug-target to narrow down the search space verification. In this work, we propose new framework, namely, Multi-Graph Regularized Nuclear Norm Minimization, predicts from three inputs: known interaction network, similarities...
Abstract While understanding molecular heterogeneity across patients underpins precision oncology, there is increasing appreciation for taking intra-tumor into account. Based on large-scale analysis of cancer omics datasets, we highlight the importance transcriptomic (ITTH) predicting clinical outcomes. Leveraging single-cell RNA-seq (scRNA-seq) with a recommender system (CaDRReS-Sc), show that heterogeneous gene-expression signatures can predict drug response high accuracy (80%). Using...
Abstract The year 2020 witnessed a heavy death toll due to COVID-19, calling for global emergency. continuous ongoing research and clinical trials paved the way vaccines. But, vaccine efficacy in long run is still questionable mutating coronavirus, which makes drug re-positioning reasonable alternative. COVID-19 has hence fast-paced treatment of its symptoms. This work builds computational models using matrix completion techniques predict drug-virus association re-positioning. aim assist...
Cellular composition and anatomical organization influence normal aberrant organ functions. Emerging spatial single-cell proteomic assays such as Image Mass Cytometry (IMC) Co-Detection by Indexing (CODEX) have facilitated the study of cellular enabling high-throughput measurement cells their localization directly in intact tissues. However, annotation cell types quantification relative tissues remain challenging. To address these unmet needs for atlas-scale datasets like Human Pancreas...
Single-cell RNA-seq has inspired new discoveries and innovation in the field of developmental cell biology for past few years is useful studying cellular responses at individual resolution. But, due to paucity starting RNA, data acquired have dropouts. To address this, we propose a deep matrix factorization-based method, deepMc, impute missing values gene expression data. For architecture our approach, draw motivation from great success learning solving various machine problems. In this...
Abstract Extensive in vitro cancer drug screening datasets have enabled scientists to identify biomarkers and develop machine learning models for predicting sensitivity. While most advancements focused on omics profiles, sensitivity scores precalculated by the original sources are often used as-is, without consideration variabilities between studies. It is well-known that significant inconsistencies exist across due differences experimental setups preprocessing methods obtain scores. As a...
This study formulates antiviral repositioning as a matrix completion problem wherein the drugs are along rows and viruses columns. The input is partially filled, with ones in positions where drug has been known to be effective against virus. curated metadata for antivirals (chemical structure pathways) (genomic symptoms) encoded into our framework graph Laplacian regularization. We then frame resulting multiple regularized (GRMC) deep factorization. solved by using novel optimization method...
This work proposes matrix completion via deep factorization on graphs. The is motivated by the success of two very recent studies (shallow) graphs and (without graphs). We show that proposed improves over both - shallow techniques factorization. Experiments are carried out challenging real-life problem modeling drug-target interactions.
In drug target interaction (DTI) the interactions of some (a subset) drugs on targets are known. The goal is to predict all targets. One approach formulate this as a matrix completion problem, where having along rows and columns partially filled. So far standard approaches such nuclear norm minimization factorization have been used address problem. work, we propose deep improve prediction results. Experiments performed benchmark databases comparison carried out with state-of-the-art...
Our paper aims to build a classification-model which delineates the typical motion-related activities performed at metro station using smart phone sensors. We focus on movements, such as climbing stairs or moving in lift, waiting security, turnstile check out and, platform while for train. Such classifier estimates crowd levels metro-station (and trains an indirect sense), thereby adding towards vision of efficient travel. However, building is challenging due non-trivial decision boundaries...
Abstract Drug discovery is an important field in the pharmaceutical industry with one of its crucial chemogenomic process being drug-target interaction prediction. This determination expensive and laborious, which brings need for alternative computational approaches could help reduce search space biological experiments. paper proposes a novel framework (DTI) prediction: Multi-Graph Regularized Deep Matrix Factorization (MGRDMF). The proposed method, motivated by success deep learning, finds...
Investigation of existing drugs is an effective alternative to the discovery new for treating diseases. This task drug re-positioning can be assisted by various kinds computational methods predict best indication a given open-source biological datasets. Owing fact that similar tend have common pathways and disease indications, association matrix assumed low-rank structure. Hence, problem drug-disease prediction modeled as completion problem. In this work, we propose novel framework makes use...
Abstract Motivation Single cell RNA sequencing has been proved to be revolutionary for its potential of zooming into complex biological systems. Genome wide expression analysis at single resolution, provides a window dynamics cellular phenotypes. This facilitates characterization transcriptional heterogeneity in normal and diseased tissues under various conditions. It also sheds light on development or emergence specific populations However, owing the paucity input RNA, typical data features...
Motivation: A fully automatic workflow for scan plane prescription is desirable in clinical settings. Goal(s): Our goal to demonstrate a deep learning-based MRI automated MR scanning the prostate. Approach: This new will identify anatomical landmarks and planes prostate planning (coverage, FOV orientation) from coil sensitivity 3plane scout images. Results: The anatomy recognition showed acceptable average location error below 5mm orientation 10 degrees. Impact: As no interaction operator...
Abstract Single cell RNA-seq has fueled discovery and innovation in medicine over the past few years is useful for studying cellular responses at individual resolution. But, due to paucity of starting RNA, data acquired highly sparse. To address this, We propose a deep matrix factorization based method, deepMc, impute missing values gene-expression data. For architecture our approach, draw motivation from great success learning solving various Machine problems. In this work, support method...
Abstract The identification of interactions between drugs and target proteins is crucial in pharmaceutical sciences. experimental validation genomic drug discovery laborious expensive; hence, there a need for efficient accurate in-silico techniques which can predict potential drug-target to narrow down the search space verification. In this work, we propose new framework, namely, Multi Graph Regularized Nuclear Norm Minimization, predicts from three inputs: known interaction network,...
Abstract The advent of single-cell open-chromatin profiling technology has facilitated the analysis heterogeneity activity regulatory regions at resolution. However, stochasticity and availability low amount relevant DNA, cause high drop-out rate noise in profiles. We introduce here a robust method called as forest imputation trees (FITs) to recover original signals from highly sparse noisy FITs makes multiple avoid bias during restoration read-count matrices. It resolves challenging issue...
Summary While understanding heterogeneity in molecular signatures across patients underpins precision oncology, there is increasing appreciation for taking intra-tumor into account. Single-cell RNA-seq (scRNA-seq) technologies have facilitated investigations the role of transcriptomic (ITTH) tumor biology and evolution, but their application to silico models drug response has not been explored. Based on large-scale analysis cancer omics datasets, we highlight utility ITTH predicting clinical...