Biology subjects

Dhruba, S. R.

Publications and source records attributed to Dhruba, S. R..

9 recordsLinked to original sources

bioRxiv2025

Deep learning inference of cell type-specific gene expression from breast tumor histopathology

Cell type-specific gene expression from single-cell RNA sequencing (RNA-seq) is valuable for breast cancer precision oncology but available cohorts are still limited due to its high cost. Deconvolution methods infer cell type-specific expression from bulk RNA-seq at a lower cost, yet expenses and processing time of bulk RNA-seq are also non-negligible and limit their application too. To address these limitations, we developed SLIDE-EX (SLide-based Inference of DEconvolved gene EXpression), a deep-learning tool that predicts cell type-specific gene expression and abundances directly from routine breast cancer histopathology whole slide images (WSIs), using deconvolved bulk RNA-seq data as training labels. Trained on the TCGA-breast cohort and tested in cross validation and on an independent cohort of 160 cases, SLIDE-EX robustly infers the expression of thousands of genes across 9 distinct cell types, performing best for cancer associated fibroblasts and cancer cells. The abundance of these two cell types could also be robustly predicted, together with that of myeloid cells. The robustly predicted genes reflect key biological functions of their respective cell types. From a translational perspective, the inferred cell type specific expression profiles predict chemotherapy response more accurately than models based on direct prediction from the slides or from the inferred bulk expression in two independent cohorts. Going forward, SLIDE-EX is a generic approach that opens up possibilities for rapid, cost-effective cell type-specific gene expression inference in potentially any cancer type, further democratizing the characterization of the tumor microenvironment.

cancer biology↗

bioRxiv2025

Single-cell-guided identification of logic-gated antigen combinations for designing effective and safe CAR therapy

Chimeric antigen receptor (CAR) T-cell therapy has revolutionized the treatment of hematological malignancies. However, its application in solid tumors remains limited because single targets are unlikely to suffice due to tumor antigen heterogeneity and off-tumor toxicities. To overcome these obstacles, we developed LogiCAR designer, a computational approach that utilizes single-cell transcriptomics data from patient tumors to systematically identify the cancer-specific antigen circuits with logic gates ("AND," "OR," and "NOT") that target the majority of cancer cells in a tumor while sparing normal cells and tissues as much as possible. LogiCAR designer efficiently scales to higher-order antigen combinations involving up to five genes. Applied to a large-scale dataset encompassing approximately 2 million cells (including > 620k tumor cells) from 342 clinical patient samples across all major breast cancer subtypes, LogiCAR designer identified antigen circuits with enhanced tumor-targeting efficacy and improved safety profiles compared to both previously reported circuits and single-target therapies in clinical trials. However, even these optimized shared circuits still proved insufficient for some patients. We hence systematically studied LogiCAR designers ability to identify highly effective CAR circuits that are individualized to each patient. Remarkably, such personalized CAR circuits provide estimated tumor-targeting efficacy tantamount to complete response in 76% of patients and partial response for all patients. Taken together, this analysis is the first systematic quantification of the efficacy and safety of all possible CAR circuits, showing that: (a) the quality of existing solutions leaves much to be desired; (b) the ability of shared circuits optimized across many patients is moderate, and finally, (c) individually tailored circuits offer significantly higher tumor-targeting efficacies for patients. LogiCAR designer offers a rigorous, data-driven way to facilitate the rational design of safe and effective CAR-based immunotherapies for cancer. Statement of Significance Development of a computational approach that efficiently identifies logic-gated CAR target combinations, called circuits, from single-cell transcriptomics, addressing a critical unmet clinical need. Application to the largest ensemble of breast cancer datasets to date, comprising [~]2 million cells (> 620k tumor cells) from 17 clinical cohorts, to identify CAR circuits predicted to be effective. Comprehensive safety profiling of candidate circuits spanning major tissues at both RNA and protein levels. Logic-gated CAR circuits generated by our pipeline address tumor heterogeneity and achieve efficacy and safety scores that surpass clinical trial and previously computationally identified circuits. Individualized rational CAR design offers a transformative approach to deliver precision-engineered CAR therapies with unprecedented efficacy.

bioinformatics↗

bioRxiv2025

Path2Omics: Enhanced transcriptomic and methylation prediction accuracy from tumor histopathology

Precision oncology is becoming increasingly integral to clinical practice, demonstrating notable improvements in treatment outcomes. While molecular data provide comprehensive insights, obtaining such data remains costly and time-consuming. To address this challenge, we developed Path2Omics, a deep learning model that predicts gene expression and methylation from histopathology for 23 cancer types. Path2Omics was trained on 20,497 slides (9,456 formalin-fixed and paraffin-embedded (FFPE) and 11,041 fresh frozen (FF)) from 8,007 patients across 23 The Cancer Genome Atlas cohorts. When tested on FFPE slides, the most readily available format in clinical pathology practice, the integrated model outperformed its individual FF and FFPE components, robustly predicting nearly 5,000 genes on average, approximately five times more than our recently published DeepPT model. Externally evaluated on seven independent cohorts, Path2Omics robustly predicted the expression of approximately 4,400 genes, yielding a 30% increase over the FFPE model alone. Finally, we demonstrate that the inferred gene expression is nearly as effective as the actual values in predicting patient survival and treatment response. These results lay the basis for using Path2Omics to advance precision oncology from histopathology slides in a speedy and cost-effective manner. Statement of significancePath2Omics is a deep learning model that accurately predicts gene expression and methylation from histopathology slides across 23 cancer types. Unlike existing approaches that rely solely on FFPE slides for training, Path2Omics leverages both FFPE and FF slides by constructing two separate models and integrating them. Downstream analyses show that the inferred values from Path2Omics are nearly as effective as actual values in predicting patient survival and treatment response.

bioinformatics↗

bioRxiv2024

IMMClock reveals immune aging and T cell function at single-cell resolution

The aging of the immune system substantially impacts individual immune responses, yet accurately quantifying immune age remains a complex challenge. Here we developed IMMClock, a novel immune aging clock that uses gene expression data to predict the biological age of individual CD8 T cells, CD4 T cells, and NK cells. The accuracy of IMMClock is first validated across multiple independent datasets, demonstrating its robustness. Second, utilizing the IMMClock, we find that intrinsic cellular aging processes are more strongly altered during immune aging than differentiation processes. Thirdly, our analysis confirms the strong associations between immune aging and established processes such as cellular senescence, exhaustion, and telomere length at the single cell level. Furthermore, immune aging is accelerated under several disease conditions such as type 2 diabetes, heart disease, and cancer. Finally, we apply IMMClock to analyze a perturb-seq gene activation screen of T cell functionality. We find that the post-perturbation immune age of individual T cells is strongly correlated with their pre-perturbation immune age. Furthermore, the immune age at resting state of individual T cells is strongly predictive of their post-stimulation activation state. Overall, IMMClock advances our understanding of immune aging by providing precise, single-cell level age estimations. Its future applications hold promise for identifying interventions that concomitantly rejuvenate and activate T cells, potentially enhancing efforts to counteract age-related immune decline.

bioinformatics↗

bioRxiv2024

ecPath detects ecDNA in tumors from histopathology images

Circular extrachromosomal DNA (ecDNA) can drive tumor initiation, progression and resistance in some of the most aggressive cancers and is emerging as a promising anti-cancer target. However, detection currently requires costly whole-genome sequencing (WGS) or labor-intensive cytogenetic or FISH imaging, limiting its application in routine clinical diagnosis. To overcome this, we developed ecPath (ecDNA from histopathology), a computational method for predicting ecDNA status from routinely available hematoxylin and eosin (H&E) images. ecPath implements a deep-learning method we call transcriptomics-guided learning, which utilizes both transcriptomics and H&E images during the training phase to enable successful ecDNA prediction from H&E images alone, a task not achievable with models trained on H&E images only. It is trained on more than 6,000 tumor whole-slide images from the TCGA cohort with the best performance in predicting ecDNA status in brain and stomach tumors (average AUC=0.78). ecPath revealed that ecDNA-positive tumors are enriched with pleomorphic, larger and high-density nuclei. Testing in an independent cohort, ecPath predicted ecDNA status of 985 pediatric brain tumor patients with an AUC of 0.72. Finally, we applied ecPath to identify ecDNA-positive tumors in the TCGA cohort for which no WGS data were available. Like WGS-based ecDNA-positive labels, the predicted ecDNA-positive status also identify poor prognoses for low grade glioma patients. These results demonstrate that ecPath enables the detection of ecDNA from routinely available H&E imaging alone and help nominate aggressive tumors with ecDNA to study and target it.

pathology↗

bioRxiv2024

The expression patterns of different cell types and their interactions in the tumor microenvironment are predictive of breast cancer patient response to neoadjuvant chemotherapy

The tumor microenvironment (TME) is a complex ecosystem of diverse cell types whose interactions govern tumor growth and clinical outcome. While multiple studies have extensively charted the TMEs impact on immunotherapy, its role in chemotherapy response remains less explored. To address this, we developed DECODEM (DEcoupling Cell-type-specific Outcomes using DEconvolution and Machine learning), a generic computational framework leveraging cellular deconvolution of bulk transcriptomics to associate gene expression of individual cell types in the TME with clinical response. Employing DECODEM to analyze gene expression of breast cancer patients treated with neoadjuvant chemotherapy across three bulk cohorts, we find that the expression of specific immune cells (myeloid, plasmablasts, B-cells) and stromal cells (endothelial, normal epithelial, CAFs) are highly predictive of chemotherapy response, achieving the same prediction levels as the expression of malignant cells. Notably, ensemble models integrating the estimated expression of different cell types perform best and outperform models built on the original tumor bulk expression. These findings and the models generalizability are further tested and validated in two single-cell (SC) cohorts of triple negative breast cancer. To investigate the possible role of immune cell-cell interactions (CCIs) in mediating chemotherapy response, we extended DECODEM to DECODEMi to identify such key functionally important CCIs, validated in SC data. Our findings highlight the importance of active pre-treatment immune infiltration for chemotherapy success. DECODEM and DECODEMi are made publicly available to facilitate studying the role of the TME in mediating response in a wide range of cancer indications and treatments.

bioinformatics↗

bioRxiv2023

Deactivation of ligand-receptor interactions enhancing lymphocyte infiltration drives melanoma resistance to Immune Checkpoint Blockade

Immune checkpoint blockade (ICB) is a promising cancer therapy; however, resistance often develops. To learn more about ICB resistance mechanisms, we developed IRIS (Immunotherapy Resistance cell-cell Interaction Scanner), a machine learning model aimed at identifying candidate ligand-receptor interactions (LRI) that are likely to mediate ICB resistance in the tumor microenvironment (TME). We developed and applied IRIS to identify resistance-mediating cell-type-specific ligand-receptor interactions by analyzing deconvolved transcriptomics data of the five largest melanoma ICB therapy cohorts. This analysis identifies a set of specific ligand-receptor pairs that are deactivated as tumors develop resistance, which we refer to as resistance deactivated interactions (RDI). Quite strikingly, the activity of these RDIs in pre-treatment samples offers a markedly stronger predictive signal for ICB therapy response compared to those that are activated as tumors develop resistance. Their predictive accuracy surpasses the state-of-the-art published transcriptomics biomarker signatures across an array of melanoma ICB datasets. Many of these RDIs are involved in chemokine signaling. Indeed, we further validate on an independent large melanoma patient cohort that their activity is associated with CD8+ T cell infiltration and enriched in hot/brisk tumors. Taken together, this study presents a new strongly predictive ICB response biomarker signature, showing that following ICB treatment resistant tumors turn inhibit lymphocyte infiltration by deactivating specific key ligand-receptor interactions.

bioinformatics↗

bioRxiv2023

Robust prediction of patient outcomes with immune checkpoint blockade therapy for cancer using common clinical, pathologic, and genomic features

Despite the revolutionary impact of immune checkpoint blockade (ICB) in cancer treatment, accurately predicting patients responses remains elusive. We analyzed eight cohorts of 2881 ICB-treated patients across 18 solid tumor types, the largest dataset to date, examining diverse clinical, pathologic, and genomic features. We developed the LOgistic Regression-based Immunotherapy-response Score (LORIS) using a transparent, compact 6-feature logistic regression model. LORIS outperforms previous signatures in ICB response prediction and can identify responsive patients, even those with low tumor mutational burden or tumor PD-L1 expression. Importantly, LORIS consistently predicts both objective responses and short-term and long-term survival across most cancer types. Moreover, LORIS showcases a near-monotonic relationship with ICB response probability and patient survival, enabling more precise patient stratification across the board. As our method is accurate, interpretable, and only utilizes a few readily measurable features, we anticipate it will help improve clinical decision-making practices in precision medicine to maximize patient benefit.

cancer biology↗

bioRxiv2022

Predicting patient treatment response and resistance via single-cell transcriptomics of their tumors

Tailoring the best treatments to cancer patients is an important open challenge. Here, we build a precision oncology data science and software framework for PERsonalized single-Cell Expression-based Planning for Treatments In Oncology (PERCEPTION). Our approach capitalizes on recently published matched bulk and single-cell transcriptome profiles of large-scale cell-line drug screens to build treatment response models from patients single-cell (SC) tumor transcriptomics. First, we show that PERCEPTION successfully predicts the response to monotherapy and combination treatments in screens performed in cancer and patient-tumor-derived primary cells based on SC-expression profiles. Second, it successfully stratifies responders to combination therapy based on the patients tumors SC-expression in two very recent multiple myeloma and breast cancer clinical trials. Thirdly, it captures the development of clinical resistance to five standard tyrosine kinase inhibitors using tumor SC-expression profiles obtained during treatment in a lung cancer patients cohort. Notably, PERCEPTION outperforms state-of-the-art bulk expression-based predictors in all three clinical cohorts. In sum, this study provides a first-of-its-kind conceptual and computational method that is predictive of response to therapy in patients, based on the clonal SC gene expression of their tumors.

bioinformatics↗