Biology subjects

Bhasin, A.

Publications and source records attributed to Bhasin, A..

7 recordsLinked to original sources

bioRxiv2026

Chemical Descriptors and Deep Learning Embeddings for Scoring de novo Peptide Designs

Peptides occupy a valuable niche between small molecules and biologics, but the clinical translation of de novo peptide designs requires rigorous scoring to simultaneously optimise target binding affinity alongside multiple developability traits, including stability, membrane permeability, aggregation propensity, and non-fouling behaviour. Here, we evaluate two distinct approaches for scoring these candidates: classical chemical descriptors and modern deep learning representations derived from protein language and folding models. Assembling nine public datasets spanning five developability traits and four binding-affinity endpoints, we find sequence-derived chemical descriptors alone contain sufficient information to predict developability task labels effectively. Given their drastically lower computational cost and higher interpretability, classical machine learning models trained on these simple descriptors frequently match or approach the performance of complex deep learning architectures, emerging as a highly efficient and interpretable alternative for high-throughput scoring. Finally, for scoring binding affinity, we demonstrate that Boltz-2 pair representations capture the most information among the tested representations; however, the model's predictive power is confounded by a significant bias from the molecular weight of the peptides. Together, these results establish a comprehensive assessment of state-of-the-art methods for predicting both peptide developability and binding affinity, highlighting the enduring value of interpretable chemical descriptors alongside deep learning in the scoring and selection of de novo peptide designs.

bioinformatics↗

bioRxiv2025

Single-cell analysis reveals a universal pericyte signature associated with poor clinical outcome and immune T cell dysfunction in thyroid cancer and other cancers

BackgroundCancer is a devastating disease with rising incidence rates and generally poor outcomes. Single-cell genomic technologies enabled the profiling of thousands of individual cells from tumors to understand their role in cancer progression. In this study, we are characterizing pericytes from tumor microenvironment single-cell data of human cancers to assess associations with patient prognosis and achieve mechanistic insights. Designwe have characterized pericytes from tumor microenvironment (TME) single-cell datasets of different cancers to assess associations with patient prognosis and achieve mechanistic insights. MethodsFor comprehensive pericyte characterization, 23 publicly available single-cell RNA sequencing (scRNA-seq) datasets corresponding to 15 cancers and 1 benign tumor were downloaded from the Gene Expression Omnibus (GEO) database and Tisch repositories. The datasets were processed using a uniform workflow, including quality control, normalization, variable gene selection, clustering, and cellular annotation (based on automated and manual markers). The annotation of the pericytes was performed based on the score calculated using our previously validated pericyte gene signature. The differential gene expression analysis between pericytes and fibroblasts in each dataset was performed to identify pericyte signature. Comparative analysis of the signatures was performed to identify a universal pericyte signature that was evaluated for pericyte specificity using peripheral blood mononuclear cells (PBMC) data. The universal pericytes signature genes were evaluated for cancer outcome associations in The Cancer Genome Atlas (TCGA) datasets using the Survival Genie Platform. Furthermore, pathway and gene-network analyses were performed to understand the biological significance of the pan-cancer universal pericyte signature. ResultsWe conducted an analysis of 23 single cell RNA-sequencing datasets from 14 cancers: breast, pancreatic, bladder, ovarian, non-small cell lung, cervical, prostate, basal cell, colon, head and neck, papillary and anaplastic thyroid cancers, as well as melanoma and lymphomas; and a benign tumor (neurofibroma). The pericytes gene set enabled the identification of a distinct pericyte cluster comprising more than 50 cells in the majority of datasets. Differential gene expression analysis between pericytes and fibroblasts identified heterogenous pericyte signatures for different cancers. Initial cell-cell communication analysis in anaplastic thyroid cancer indicated that pericytes are highly communicative cells, ranking among the top ligand-secreting populations, and actively engaging with anaplastic thyroid cancer cells and T cells. A comparative analysis of pericyte signatures generated from different cancers yielded a core signature of 100 genes that were consistently overexpressed in more than 60% of the cancer datasets. Notably, 78% of these genes were not expressed in PBMCs, supporting their pericyte specificity. Survival analysis using TCGA datasets identified a 19 genes pan-cancer pericyte signature associated with poor prognosis (HR>1 in at least 30% datasets). This pan-cancer pericyte signature included genes such as TPM2 and EHD2, whose overexpression was significantly associated with poor overall survival across multiple aggressive cancers, including bladder, head and neck, lung, pancreatic, skin, and thyroid cancers. Interestingly, many of these genes showed a strong positive correlation (e.g. PDGFRB) with T cell exhaustion-related genes in the TCGA pan-cancer cohort (n=10,293 samples), suggesting a potential mechanistic link between pericytes and T cell exhaustion. Importantly, we have also identified a unique tyrosine kinase (TK) receptors gene signature in pericytes compared to fibroblasts or other cell types by heatmap analysis across the different cancers. ConclusionsThis large-scale scRNA-seq analysis of multiple datasets defines a pan-cancer pericyte signature and sheds light on pericyte activity, particularly their communication with T cells, suggesting a role in T cell exhaustion, a key factor in the failure of many cancer therapies. This study underscores the pivotal role of pericyte population enhancement as a critical regulator of the TME, facilitating immune evasion and driving disease progression. TK expression in pericytes may serve as a predictive biomarker for TKI responsiveness, offering a cellular target whose expression pattern may guide therapeutic decisions and improve patient stratification in TKI-based treatment regimens. Translational significanceThese findings provide clinically actionable insights by identifying pericyte-derived molecular targets that may refine patient risk stratification and support personalized treatment approaches. The pericyte-associated markers uncovered here show strong prognostic value across multiple cancer types, highlighting their potential to improve outcome prediction. Moreover, the specificity of these targets positions them as promising candidates for therapeutic intervention and for incorporation into future biomarker-driven clinical trials.

cancer biology↗

bioRxiv2025

Exploring the Conformational Landscape of Adenylate Kinase and Beyond: A Benchmark of Protein Folding Models

Protein folding models have revolutionized structure prediction but struggle to capture conformational flexibility. Recent studies perturb inputs or parameters to sample alternative conformations, while diffusion-based approaches generate conformational ensembles directly. Although the former have been benchmarked to some extent, the latter have yet to be evaluated, and sub-domain dynamics validation remains limited. Here, we present a systematic benchmark of nine methods across 20 monomeric proteins with active and inactive states. We extend the pairwise aligned error metric to ensembles and reveal that protein identity exerts a non-negligible influence on model performance. Focusing on Adenylate Kinase, a well-studied enzyme with extensive molecular dynamics (MD) data, we find that Chai-1 performs the best in recovering known conformations, identifying mobile regions, and capturing transition trajectories. These results highlight the potential of generative models as efficient alternatives to MD for exploring protein conformational dynamics and provide a rigorous benchmark for dynamic structure prediction.

bioinformatics↗

bioRxiv2025

Whole-body Loss of FSP27 Impairs Cognitive Function via Disruption of Neuro-Metabolic Pathways

Fat-Specific Protein 27 (FSP27), originally identified for its role in adipocyte lipid metabolism and energy homeostasis, plays a key role in regulating lipolysis and maintaining insulin sensitivity. Beyond adipose tissue, emerging studies have uncovered its involvement in hepatic and skeletal muscle function. More recently, FSP27 has also been implicated in maintaining vascular health through its influence on endothelial signaling. Despite growing insights into FSP27s systemic functions, its involvement in the central nervous system and cognitive regulation have remained unexplored. In this study, we present the first evidence that FSP27 is a critical regulator of cognitive function. Utilizing a global FSP27 knockout (Fsp27-/-) mouse model, we demonstrate that FSP27 deficiency results in significant impairments in learning, memory retention, and spatial awareness. To elucidate the underlying molecular mechanisms, genome-wide transcriptome profiling of brain tissue from Fsp27-/- mice was performed, revealing significant (P<0.05) alterations in gene expression related to neurocognition and metabolic pathways. Notably, FSP27 deletion was associated with genomic instability (P=0.05), downregulation of genes essential for axonal transport (NES=-1.91, FDR q-value=0.21), neuronal plasticity, and brain development (NES=-1.91, FDR q-value=0.28), along with signatures of disrupted systemic metabolism and elevated stress responses (NES=1.81, FDR q-value=0.20) in the brain, the processes tightly linked to cognitive dysfunction. Collectively, these findings establish FSP27 as a molecular node connecting metabolic regulation with cognitive health and identify it as a promising target for therapeutic intervention in neurodegenerative and cognitive disorders. Significance StatementFSP27, previously known for its role in lipid metabolism, is now revealed to be a critical regulator of brain function. Our study provides the first direct evidence that FSP27 supports learning and memory by maintaining neuronal integrity and regulating neurocognitive gene networks. These findings uncover an unexpected metabolic- cognitive link and position FSP27 as a promising molecular target for therapeutic strategies aimed at preventing or treating neurodegenerative and cognitive disorders.

animal behavior and cognition↗

bioRxiv2025

CoV-UniBind: A Unified Antibody Binding Database for SARS-CoV-2

Since the emergence of SARS-CoV-2, numerous studies have investigated antibody interactions with viral variants in vitro, and several datasets have been curated to compile available protein structures and experimental measurements. However, existing data remain fragmented, limiting their utility for the development and validation of machine learning models for antibody-antigen interaction prediction. Here, we present CoV-UniBind, a unified database comprising over 75,000 entries of SARS-CoV-2 antibody-antigen sequence, binding, and structural data, integrated and standardised from three public sources and multiple peer-reviewed publications. To demonstrate its utility, we benchmarked multiple protein folding and inverse folding models across tasks relevant to antibody design and vaccine development. We expect CoV-UniBind to facilitate future computational efforts in antibody and vaccine development against SARS-CoV-2. Availability and implementationThe curated datasets, structures, model scores and antibody synonyms are free to download at https://huggingface.co/datasets/InstaDeepAI/cov-unibind. Folded structures are available upon request.

immunology↗

bioRxiv2025

Rapid enrichment of progenitor exhausted neoantigen-specific CD8 T cells from peripheral blood

Neoantigen-reactive peripheral blood lymphocytes (NeoPBL) are tumor-specific T cells found at ultra-low frequencies in the blood. Unlike tumor-infiltrating lymphocytes (TIL), NeoPBL exist in a favorable less dysfunctional phenotypic state in vivo, but their rarity has precluded their effective use as cell therapy. Leveraging a priori knowledge of bona fide neoantigens, we combined high-intensity neoantigen stimulation with bead extraction of neoantigen peptide-pulsed target cells to enable the enrichment of NeoPBL to frequencies comparable to ex vivo cultured TIL over a 28-day period. Throughout this process, NeoPBL demonstrate specific reactivity against autologous tumor organoids and maintain memory-like features, including elevated expression of CD28 and TCF7. We additionally demonstrate that NeoPBL reactivity is polyclonal, encompassing multiple clonotypes that are detectable within in vivo TIL populations, underscoring physiological specificity for the targeted neoantigens. This streamlined process yields clinically relevant cell doses and enables identification and expansion of blood-derived neoantigen-specific TCRs. By potentially avoiding additional surgical risks and protracted delays of TIL and individualized TCR-engineered methods, the NeoPBL platform may have clinical and practical advantages. Ultimately, NeoPBL combines intrinsic cell fitness, minimal invasiveness and rapidity to potentially facilitate personalized adoptive cell therapy for cancer.

immunology↗

bioRxiv2024

The shared genetic basis of leaf morphology and tensile resistance underlies the effect of growing season length in a widespread perennial grass

PremiseLeaf tensile resistance, a leafs ability to withstand pulling forces, is an important determinant of plant ecological strategies. One potential driver of leaf tensile resistance is growing season length. When growing seasons are long, strong leaves--which often require more time and resources to construct than weak leaves--may be more advantageous than when growing seasons are short. Growing season length and other ecological conditions may also impact the morphological traits that underlie leaf tensile resistance. MethodsTo understand variation in leaf tensile resistance, we measured size-dependent leaf strength and size-independent leaf toughness in diverse genotypes of the widespread perennial grass Panicum virgatum (switchgrass) in a common garden. We then used quantitative genetic approaches to estimate the heritability of leaf tensile resistance and whether there were genetic correlations between leaf tensile resistance and other morphological traits. Key ResultsLeaf tensile resistance was positively associated with aboveground biomass (a proxy for fitness). Moreover, both measures of leaf tensile resistance exhibited high heritability and were positively genetically correlated with leaf lamina thickness and leaf mass per area (LMA). Leaf tensile resistance also increased with habitat-of-origin growing season length and this effect was mediated by both LMA and leaf thickness. ConclusionsDifferences in growing season length may promote selection for different leaf lifespans and may explain existing variation in leaf tensile resistance in P. virgatum. In addition, the high heritability of leaf tensile resistance suggests that P. virgatum will be able to respond to climate change as growing seasons lengthen.

plant biology↗