Biology subjects

Sobhan, M.

Publications and source records attributed to Sobhan, M..

9 recordsLinked to original sources

bioRxiv2025

PathPCNet: Pathway Principal Component-Based Interpretable Framework for Drug Sensitivity Prediction

Precision medicine aims to identify significant biomarkers and effective drugs tailored to individual genomic profiles, thereby enabling personalized treatment strategies. Drug efficacy is often attributed to drug response, commonly measured as the concentration of a drug required to inhibit a biological activity. In contrast, drug sensitivity reflects how strongly a tumor responds to a drug, where a lower effective dose indicates higher sensitivity. Machine learning-based drug response prediction has the potential to accelerate biomarker discovery and facilitate the development of more effective therapeutics. In this study, we present PathPCNet, a novel interpretable deep learning framework that integrates multi-omics data (copy number variation, mutation, and RNA sequencing) fused with biological pathways, drug molecular structure, and Principal Component Analysis for drug response prediction. Our model achieves a Pearson correlation coefficient of 0.941 and an R-squared of 0.885, outperforming the existing pathway-based approaches. We employ SHAP-based model interpretation to quantify the contributions of omics and drug features, uncovering key pathways and gene-drug interactions associated with resistance mechanisms. These results demonstrate the utility of integrative deep learning models not only for accurate prediction but also for generating biologically meaningful insights, which can advance drug discovery and precision oncology. In addition, the framework also facilitates the identification of important pathways, genes, and atomic attributes of drugs related to drug sensitivity and different cancer types.

bioinformatics↗

bioRxiv2025

PathX-CNN: An Enhanced Explainable Convolutional Neural Network for Survival Prediction and Pathway Analysis in Glioblastoma

MotivationConvolutional neural networks (CNNs) offer potential for analyzing non-grid structured data, such as biological array data, by converting it into image-like formats using principal component analysis (PCA) of pathway genes. However, PCA-derived principal components (PCs) from the entire dataset capture global variance but fail to extract sub-cohort (class-specific) variances. Consequently, CNNs trained on global PCs perform poorly in survival prediction of glioblastoma multiforme (GBM), and the corresponding explanation of CNN outcomes may not align with disease-relevant pathways. ResultsWe present PathX-CNN, an explainable CNN framework that addresses these limitations by integrating multi-omics data through pathway-based images derived from sub-cohort-specific PCs. PathX-CNN outperformed existing pathway-based methods in predicting long-term survival (LTS) versus non-LTS in GBM. By leveraging SHAP (SHapley Additive exPlanations), a cooperative game theory-based explainable AI method, PathX-CNN identified biologically plausible pathways associated with GBM survival. Additionally, experiments on other cancer types demonstrated superior performance compared to traditional approaches. PathX-CNN demonstrates the potential of CNNs for multiomics integration, offering both improved prediction accuracy and pathway-specific insights into disease mechanisms. Code Availabilityhttps://github.com/codebysobhan/PathX-CNN Contactamondal@fiu.edu Supplementary informationSupplementary data are available at Bioinformatics online.

cancer biology↗

bioRxiv2025

Interpreting Lung Cancer Health Disparity at Transcriptome Level

Lung cancer is a leading cause of cancer-related mortality, with disparities in incidence and outcomes observed across different racial and sex groups. Understanding the genetic factors of these disparities is critical for developing targeted treatment therapies. This study aims to identify both patient-specific and cohort-specific biomarker genes that contribute to lung cancer health disparities among African American males (AAMs), European American males (EAMs), African American females (AAFs), and European American females (EAFs). The real-world data is highly imbalanced with respect to race, and the lung cancer dataset is no exception. So, classification with race labels will generate highly biased results toward the larger cohort. We developed a computational framework by designing the classification problems with disease conditions instead of races and leveraging the local interpretability of explainable AI, SHAP (SHapley Additive exPlanations). This study used three disease conditions of lung cancer, including Lung Adenocarcinoma (LUAD), Lung Squamous Cell Carcinoma (LUSC), and Healthy samples (HEALTHY) to design four classification tasks: one 3-class problem (LUAD-LUSC-HEALTHY) and three 2-class problems (LUAD-LUSC, LUAD-HEALTHY, and LUSC-HEALTHY). This multiple-classification approach allows a LUAD patient to be interrogated via three classification problems, namely LUAD-LUSC-HEALTHY, LUAD-LUSC, and LUAD-HEALTHY, thus providing a robust approach of retrieving disparity information for individual patients through the local interpretation of SHAP. The proposed method successfully discovered the sets of genes and pathways related to health disparities in lung cancer between two cohorts, including AAMs vs. EAMs, AAFs vs. EAFs, AAMs vs. AAFs, and EAMs vs. EAFs. The discovered list of genes and pathways provide a short list for biological scientists to conduct wet lab experiment.

cancer biology↗

bioRxiv2023

MOGAT: An Improved Multi-Omics Integration Framework Using Graph Attention Networks

Integration of multi-omics data holds great promise for understanding the complex biology of diseases, particularly Alzheimers, Parkinsons, and cancer. However, the integration is challenging due to the high dimensionality and complexity of the data. Traditional machine learning methods are not well-suited for handling the complex relationships between different types of omics data. Many models were proposed that utilize graph-based learning models to extract hidden representations and network structures from different omics data to enhance cancer prediction, patient categorization, etc. The existing graph neural network-based (GNN-based) multi-omics approaches for cancer subtype prediction have three shortcomings: (a) Do not consider all types of omics data, (b) Fail to determine the relative significance of the neighboring nodes (in this case, samples or patients) when it comes to downstream analyses, such as subtype classification, patient stratification, etc., and (c) Use the same approach for generating initial graphs for different omics data. To overcome these shortcomings, we present MOGAT, a novel multi-omics integration approach, leveraging a graph attention network (GAT) model that incorporates graph-based learning with an attention mechanism. MOGAT utilizes a multi-head attention mechanism that can more efficiently extract information for a specific sample by assigning unique attention coefficients to its neighboring samples. To evaluate the performance of MOGAT, we explored its capability via a case study of predicting breast cancer subtypes. Our results showed that MOGAT performs better than the state-of-the-art multi-omics integration approaches.

bioinformatics↗

bioRxiv2022

Explainable Machine Learning to Identify Patient-specific Biomarkers for Lung Cancer

BackgroundLung cancer is the leading cause of death compared to other cancers in the USA. The overall survival rate of lung cancer is not satisfactory even though there are cutting-edge treatment methods for cancers. Genomic profiling and biomarker gene identification of lung cancer patients may play a role in the therapeutics of lung cancer patients. The biomarker genes identified by most of the existing methods (statistical and machine learning based) belong to the whole cohort or population. That is why different people with the same disease get the same kind of treatment, but results in different outcomes in terms of success and side effects. So, the identification of biomarker genes for individual patients is very crucial for finding efficacious therapeutics leading to precision medicine. MethodsIn this study, we propose a pipeline to identify lung cancer class-specific and patient-specific key genes which may help formulate effective therapies for lung cancer patients. We have used expression profiles of two types of lung cancers, lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), and Healthy lung tissues to identify LUAD- and LUSC-specific (class-specific) and individual patient-specific key genes using an explainable machine learning approach, SHaphley Additive ExPlanations (SHAP). This approach provides scores for each of the genes for individual patients which tells us the attribution of each feature (gene) for each sample (patient). ResultIn this study, we applied two variations of SHAP - tree explainer and gradient explainer for which tree-based classifier, XGBoost, and deep learning-based classifier, convolutional neural network (CNN) were used as classification algorithms, respectively. Our results showed that the proposed approach successfully identified class-specific (LUAD, LUSC, and Healthy) and patient-specific key genes based on the SHAP scores. ConclusionThis study demonstrated a pipeline to identify cohort-based and patient-specific biomarker genes by incorporating an explainable machine learning technique, SHAP. The patient-specific genes identified using SHAP scores may provide biological and clinical insights into the patients diagnosis.

cancer biology↗

bioRxiv2022

EMoMiS: A Pipeline for Epitope-based Molecular Mimicry Search in Protein Structures with Applications to SARS-CoV-2

MotivationEpitope-based molecular mimicry occurs when an antibody cross-reacts with two different antigens due to structural and chemical similarities. Molecular mimicry between proteins from two viruses can lead to beneficial cross-protection when the antibodies produced by exposure to one also react with the other. On the other hand, mimicry between a protein from a pathogen and a human protein can lead to auto-immune disorders if the antibodies resulting from exposure to the virus end up interacting with host proteins. While cross-protection can suggest the possible reuse of vaccines developed for other pathogens, cross-reaction with host proteins may explain side effects. There are no computational tools available to date for a large-scale search of antibody cross-reactivity. ResultsWe present a comprehensive Epitope-based Molecular Mimicry Search (EMoMiS) pipeline for computational molecular mimicry searches. EMoMiS, when applied to the SARS-CoV-2 Spike protein, identified eight examples of molecular mimicry with viral and human proteins. These findings provide possible explanations for (a) differential severity of COVID-19 caused by cross-protection due to prior vaccinations and/or exposure to other viruses, and (b) commonly seen COVID-19 side effects such as thrombocytopenia and thrombophilia. Our findings are supported by previously reported research but need validation with laboratory experiments. The developed pipeline is generic and can be applied to find mimicry for novel pathogens. It has applications in improving vaccine design. AvailabilityThe developed Epitope-based Molecular Mimicry Search Pipeline (EMoMiS) is available from https://biorg.cs.fiu.edu/emomis/. Contactgiri@cs.fiu.edu

bioinformatics↗

bioRxiv2021

Quantifying Intratumor Heterogeneity by Key Genes Selected Using Concrete Autoencoder

The tumor cell population in cancer tissue has distinct molecular characteristics and exhibits different phenotypes, thus, resulting in different subpopulations. This phenomenon is known as Intratumor Heterogeneity (ITH), a major contributor to drug resistance, poor prognosis, etc. Therefore, quantifying the levels of ITH in cancer patients is essential, and many algorithms do so in different ways, using different types of omics data. DEPTH (Deviating gene Expression Profiling Tumor Heterogeneity) is the latest algorithm that uses transcriptomic data to evaluate the ITH score. It shows promising performance, has strong similarity with six other algorithms and has an advantage over two algorithms that uses the same type of data (tITH, sITH). However, it has a major drawback since it uses expression values of all the genes ([~]20K genes) in quantifying ITH levels. We hypothesize that a subset of key genes is sufficient to quantify the ITH level. To prove our hypothesis, we developed a deep learning-based computational framework using unsupervised Concrete Autoencoder (CAE) to select a set of cancer-specific key genes that can be used to evaluate the ITH score. For the experiment, we used gene expression profile data of tumor cohorts of breast, kidney, and lung cancer from the TCGA repository. Using multi-run CAE, we selected three sets of key genes, each set related to breast, kidney, and lung tumor cohorts. For the three cancers stated and three molecular subtypes of lung cancer, we calculated the ITH level using all genes and key genes selected by CAE and performed a side-by-side comparison. We could reach similar conclusions for survival and prognostic outcomes based on ITH scores derived from all genes and the sets of key genes. Additionally, for subtypes of lung cancer, the comparative distribution of ITH scores derived from all and key genes remains similar. Based on these observations, it can be stated that a subset of key genes, instead of all genes, is sufficient for ITH quantification. Our results also showed that many key genes are prognostically significant, which can be used as possible therapeutic targets.

bioinformatics↗

bioRxiv2021

Molecular mimicry between Spike and human thrombopoietin may induce thrombocytopenia in COVID-19

SARS-CoV-2 causes COVID-19, a disease curiously resulting in varied symptoms and outcomes, ranging from asymptomatic to fatal. Autoimmunity due to cross-reacting antibodies resulting from molecular mimicry between viral antigens and host proteins may provide an explanation. We computationally investigated molecular mimicry between SARS-CoV-2 Spike and known epitopes. We discovered molecular mimicry hotspots in Spike and highlight two examples with tentative autoimmune potential and implications for understanding COVID-19 complications. We show that a TQLPP motif in Spike and thrombopoietin shares similar antibody binding properties. Antibodies cross-reacting with thrombopoietin may induce thrombocytopenia, a condition observed in COVID-19 patients. Another motif, ELDKY, is shared in multiple human proteins such as PRKG1 and tropomyosin. Antibodies cross-reacting with PRKG1 and tropomyosin may cause known COVID-19 complications such as blood-clotting disorders and cardiac disease, respectively. Our findings illuminate COVID-19 pathogenesis and highlight the importance of considering autoimmune potential when developing therapeutic interventions to reduce adverse reactions.

bioinformatics↗

bioRxiv2021

Multi-run Concrete Autoencoder to Identify Prognostic lncRNAs for 12 Cancers

Long non-coding RNA plays a vital role in changing the expression profiles of various target genes that leads to cancer development. Thus, identifying prognostic lncRNAs related to different cancers might help in developing cancer therapy. To discover the critical lncRNAs that can identify the origin of different cancers, we proposed to use the state-of-the-art deep learning algorithm Concrete Autoencoder (CAE) in an unsupervised setting, which efficiently identifies a subset of the most informative features. However, CAE does not identify reproducible features in different runs due to its stochastic nature. We proposed a multi-run CAE (mrCAE) to identify a stable set of features to address this issue. The assumption is that a feature appearing in multiple runs carries more meaningful information about the data under consideration. The genome-wide lncRNA expression profiles of 12 different types of cancers, a total of 4,768 samples available in The Cancer Genome Atlas (TCGA), were analyzed to discover the key lncRNAs. The lncRNAs identified by multiple runs of CAE were added to the final list of key lncRNAs, which are capable of identifying 12 different cancers. Our results showed that mrCAE performs better in feature selection than single-run CAE, standard autoencoder (AE), and other state-of-the-art feature selection techniques. This study discovered a set of top-ranking 128 lncRNAs that could identify the origin of 12 different cancers with an accuracy of 95%. Survival analysis showed that 76 of 128 lncRNAs have the prognostic capability in differentiating high- and low-risk groups of patients in different cancers. The proposed mrCAE outperformed AE, which select latent features and were thought to be the best tools for dimensionality reduction. By selecting actual features instead of pseudo-features, mrCAE can be valuable for precision medicine. The identified prognostic lncRNAs (this work), mRNAs, miRNAs, and DNA methylated genes (future work) can lead to biomarkers and therapies for different cancers.

bioinformatics↗