Biology subjects

Kosaraju, S. C.

Publications and source records attributed to Kosaraju, S. C..

4 recordsLinked to original sources

bioRxiv2026

Fairness-aware Supervised Hierarchical Contrastive Semantic Learning for Sexual Dimorphism Analysis

MotivationSexual dimorphism is a fundamental biological determinant driving systematic differences in disease susceptibility, progression, and clinical outcomes. However, current AI-based genomic models often exhibit algorithmic bias and fail to capture these sex-specific mechanisms, creating a critical barrier to unbiased precision medicine. Ensuring fairness in the context of sexual dimorphism requires understanding and addressing the distinct biological mechanisms functioning in each sex, rather than focusing solely on equalizing predictive performance. ResultsWe propose a fairness-aware supervised hierarchical contrastive learning approach, called FairHICON, to discover unbiased sex-common and sex-specific genomic drivers. Evaluations on cancer and asthma transcriptomic datasets demonstrate that FairHICON significantly outperforms state-of-the-art benchmarks, improving predictive performance by up to 9% while effectively reducing the performance gap between male and female cohorts. Furthermore, prognostic validation confirms that the identified sex-specific pathways stratify patient survival significantly better within their corresponding sex groups. This validates FairHICON to elucidate the molecular heterogeneity of sexual dimorphism, advancing inclusive precision medicine. Availability and implementationThe source code and data is available at https://github.com/datax-lab/FairHICON.

bioinformatics↗

bioRxiv2026

PIMO: Pathway-based Interpretable Multi-Omics interactions for multi-omics integration

MotivationModeling inter-omics interactions across multiple molecular levels is critical for deciphering the mechanisms underlying complex diseases. Epigenomic and structural alterations, such as DNA methylation and copy number alterations, modulate gene expression and collectively influence disease progression and patient survival outcomes. Despite advancements in deep learning-based multi-omics analysis, gene-level interactions of inter-omics have been seldom considered, due to combinational complexity and power, which limits interpretability and mechanistic insight. ResultsWe propose a Pathway-based Interpretable deep learning Multi-Omics interaction model, PIMO, that explicitly captures regulatory effects across omics layers. Experiments on multiple TCGA cancer datasets showed that PIMO consistently outperformed state-of-the-art baselines in survival analysis, up to 13% increase in the C-index. PIMO provides biologically interpretable analyses that identify important pathways, genes, and inter-omics interactions with DNA methylation and copy number alterations. Availability and implementationThe source code and data is available at https://github.com/datax-lab/PIMO.

bioinformatics↗

bioRxiv2026

CholBindNet: Interpretable Neural Networks for Cholesterol Binding Site Prediction

Cholesterol is a key modulator of membrane protein structure and function, yet predicting cholesterol binding sites remains challenging due to its undrug-like physicochemical properties. Here, we curated more than 800 high-resolution transmembrane protein structures containing cholesterol and developed an interpretable, atom-based deep-learning framework, CholBindNet, comprising four model architectures: a 3D convolutional neural network, a graph neural network, a graph attention network, and a graph convolutional network. A Positive-Unlabeled (PU) training strategy was employed to address the scarcity of true negative samples resulting from the promiscuous nature of cholesterol binding. We show that CholBindNet substantially outperforms existing deep-learning models trained on general ligand-binding datasets. The performance and generalizability of the model were further demonstrated by rapidly assessing strong, median, and weak cholesterol-binding sites in the PIEZO2 ion channel in excellent agreement with computationally expensive all-atom molecular dynamics (MD) simulations. Additionally, strong model interpretability was achieved for CholBindNet through atom-level feature encoding, Grad-CAM visualization, and attention-based scoring analysis. Overall, CholBindNet provides an efficient and scalable approach for predicting cholesterol binding sites on membrane proteins, achieving performance comparable to MD simulations while offering mechanistic biophysical insights beyond amino-acid sequence. This work hence lays the foundation for future development of deep-learning models targeting membrane protein drug-binding sites and cholesterol-modulated therapeutics. Significance StatementDeep-learning models for ligand-binding prediction have advanced rapidly, yet those trained on soluble proteins perform poorly for membrane proteins, particularly for cholesterol binding. We introduce CholBindNet, a set of neural network models specifically designed to identify cholesterol-binding sites in transmembrane proteins. CholBindNet substantially outperforms existing deep-learning approaches and accurately ranks strong, intermediate, and weak cholesterol-binding sites in close agreement with computationally intensive all-atom molecular dynamics simulations. This work provides a practical and scalable alternative to long-timescale simulations for studying cholesterol-protein interactions. The curated cholesterol benchmark and open-source models will enable broader adoption of deep learning for investigating lipid regulation of membrane proteins and for guiding the design of drugs targeting membrane-embedded binding sites.

biophysics↗

bioRxiv2025

Prediction of bacterial protein-compound interactions with only positive samples

Prediction of Compound-Protein Interactions (CPI) in bacteria is crucial to advance various pharmaceutical and chemical engineering fields, including bio-catalysis, drug discovery, and industrial processing. However, current CPI models cannot be applied for bacterial CPI prediction due to the lack of curated negative interaction samples. This paper introduces a novel Positive-Unlabeled (PU) learning framework, named BIN-PU, to address this limitation. BIN-PU generates pseudo positive and negative labels from known positive interaction data, enabling effective training of deep learning models for CPI prediction. We also propose a weighted positive loss function that weights to truly positive samples. We have validated BIN-PU with multiple CPI backbone models, comparing the performance with the existing PU model using bacterial cytochrome P450 (CYP) data. Extensive experiments demonstrate the superiority of BIN-PU over the benchmark model in predicting CPIs with only truly positive samples. Furthermore, we have validated BIN-PU on additional bacterial proteins obtained from literature review, human CYP datasets, and uncurated data for its reproducibility. We have also validated the CPI prediction for the uncurated CYP data with biological and biophysical experiments. BIN-PU represents a significant advancement in CPI prediction for bacterial proteins, opening new possibilities for improving predictive models in related biological interaction tasks.

bioinformatics↗