Biology subjects

Narayan, V.

Publications and source records attributed to Narayan, V..

7 recordsLinked to original sources

bioRxiv2026

MicrobeRT: A Framework and Suite of Fine-tuned Language Models for Genomic Sequences

Accurate and reproducible metagenomic analysis is essential for effective infectious disease surveillance. Targeted diagnostics such as culturing, antigen assays, and multiplex PCR panels identify only predefined organisms and can fail to identify atypical or novel infections. Shotgun metagenomics offers hypothesis-free detection, but adoption is limited by workflow complexity, dependence on mutable reference databases, and high computational cost. To combat these challenges, we introduce MicrobeRT, an integrated suite of genomic language models fine-tuned to classify metagenomic sequencing reads across three tasks: (i) species level taxonomic identification, (ii) antibiotic resistance gene detection, and (iii) pathogenicity assessment. The framework builds on three pre-trained genomic language models - Nucleotide Transformer (Dalla-Torre et al 2025), DNABERT (Ji et al 2021, Zhou et al 2024), and HyenaDNA (Nguyen et al 2024) - leveraging their large-scale genomic pre-training, adaptability to downstream tasks, and, in the case of HyenaDNA, ability to capture long-range sequence dependencies. The models are fine-tuned on large, curated datasets spanning microbial, viral, fungal, and human genomes, including more than 500,000 genomes and 250,000 antibiotic resistance gene sequences. To evaluate the performance of our fine-tuned genomic language models across classification tasks, we developed a benchmarking framework that compares MicrobeRT against state-of-the-art (SOTA) bioinformatics tools. Across both simulated and empirical datasets, the models achieve strong performance while reducing computational overhead. MicrobeRT is released as open-source software, enabling further fine-tuning and reproducible integration of alternative models and classifiers. We also demonstrate seamless, drop-in integration by incorporating our fine-tuned taxonomic classification models as an optional add-on within the open-source TaxTriage pipeline (Merritt et al 2026). MicrobeRT demonstrates the feasibility of three clinically relevant multi-label analyses, offering a practical path to scalable, reproducible metagenomic diagnostics and enabling broader deployment of AI/ML-enhanced infectious disease testing.

bioinformatics↗

bioRxiv2025

Evaluating the Effectiveness of Parameter-Efficient Fine-Tuning in Genomic Classification Tasks

Foundation models are increasingly being leveraged for biological tasks. To address the high memory requirements of fine-tuning large pre-trained language models, parameter efficient fine-tuning (PEFT) methods are also being increasingly utilized. Previous studies have shown minimal, if any, loss in performance when using PEFT on binary classification tasks. However, the impact of using PEFT on tasks with large classification spaces has not been systemically evaluated. In this work, we apply PEFT to the problem of taxonomic classification using pre-trained genomic language models as the classification backbone. We explore various training strategies--including PEFT, full fine-tuning, and partial fine-tuning--for classifying sequences at the superkingdom, phylum, and genus levels. We find that PEFT-trained models significantly underperform compared to those trained via full fine-tuning or partial fine-tuning. Additionally, we demonstrate increased performance of pretrained models over those randomly initialized.

bioinformatics↗

bioRxiv2024

The PRC2.1 Subcomplex Opposes G1 Progression through Regulation of CCND1 and CCND2

Progression through the G1 phase of the cell cycle is the most highly regulated step in cellular division. We employed a chemogenetic approach to discover novel cellular networks that regulate cell cycle progression. This approach uncovered functional clusters of genes that altered sensitivity of cells to inhibitors of the G1/S transition. Mutation of components of the Polycomb Repressor Complex 2 rescued proliferation inhibition caused by the CDK4/6 inhibitor palbociclib, but not to inhibitors of S phase or mitosis. In addition to its core catalytic subunits, mutation of the PRC2.1 accessory protein MTF2, but not the PRC2.2 protein JARID2, rendered cells resistant to palbociclib treatment. We found that PRC2.1 (MTF2), but not PRC2.2 (JARID2), was critical for promoting H3K27me3 deposition at CpG islands genome-wide and in promoters. This included the CpG islands in the promoter of the CDK4/6 cyclins CCND1 and CCND2, and loss of MTF2 lead to upregulation of both CCND1 and CCND2. Our results demonstrate a role for PRC2.1, but not PRC2.2, in antagonizing G1 progression in a diversity of cell linages, including CML, breast cancer and immortalized cell lines.

molecular biology↗

bioRxiv2023

High throughput PRIME editing screens identify functional DNA variants in the human genome

Despite tremendous progress in detecting DNA variants associated with human disease, interpreting their functional impact in a high-throughput and base-pair resolution manner remains challenging. Here, we develop a novel pooled prime editing screen method, PRIME, which can be applied to characterize thousands of coding and non-coding variants in a single experiment with high reproducibility. To showcase its applications, we first identified essential nucleotides for a 716 bp MYC enhancer via PRIME-mediated saturation mutagenesis. Next, we applied PRIME to functionally characterize 1,304 non-coding variants associated with breast cancer and 3,699 variants from ClinVar. We discovered that 103 non-coding variants and 156 variants of uncertain significance are functional via affecting cell fitness. Collectively, we demonstrate PRIME capable of characterizing genetic variants at base-pair resolution and scale, advancing accurate genome annotation for disease risk prediction, diagnosis, and therapeutic target identification.

genomics↗

bioRxiv2023

An Enterobacteriaceae bloom in aging animals is restrained by the gut microbiome

The gut microbiome plays important roles in host function and health. Core microbiomes have been described for different species, and imbalances in their composition, known as dysbiosis, are associated with pathology. Changes in the gut microbiome and dysbiosis are common in aging, possibly due to multi-tissue deterioration, which includes metabolic shifts, dysregulated immunity, and disrupted epithelial barriers. However, the characteristics of these changes, as reported in different studies, are varied and sometimes conflicting. Using clonal populations of C. elegans to highlight trends shared among individuals, and employing NextGen sequencing, CFU counts and fluorescent imaging to characterize age-dependent changes in worms raised in different microbial environments, we identified an Enterobacteriaceae bloom as a common denominator in aging animals. Experiments using Enterobacter hormachei, a representative commensal, suggested that the Enterobacteriaceae bloom was facilitated by a decline in Sma/BMP immune signaling in aging animals and demonstrated its detrimental potential for increasing susceptibility to infection. However, such detrimental effects were context-dependent, mitigated by competition with commensal communities, highlighting the latter as determinants of healthy versus unhealthy aging, depending on their ability to restrain opportunistic pathobionts.

physiology↗

bioRxiv2022

Misaligned plastic and evolutionary responses of lifespan to novel carbohydrate diets

Diet elicits varied effects on longevity across a wide range of animal species. For example, diets low in protein and high in carbohydrate typically extend lifespan while diets high in protein tend to reduce it. Although studies have also shown that diet-induced lifespan changes can persist through transgenerational plasticity, whether such changes lead to evolutionary shifts in lifespan remains unclear. In this study we combine experimental evolution and phenotypic plasticity assays to address this gap. Using Drosophila serrata, we investigated the evolutionary potential of lifespan in response to four novel diets spanning a carbohydrate-protein gradient. We also examined developmental plasticity effects using a set of control populations that were raised on the four novel environments. Our results show that although lifespan evolved in response to changes in dietary carbohydrate concentration, the plastic responses for lifespan differed from the evolved responses. The direction of the evolved response (increased lifespan) observed on low carbohydrate diets was in the opposite direction to the plastic response (decreased lifespan). Our results imply that plastic responses to low carbohydrates can be maladaptive for lifespan and misaligned with the evolved responses, laying the groundwork for future investigations of carbohydrate contributions to evolved and plastic effects on lifespan.

evolutionary biology↗

bioRxiv2020

Precise Estimation of In Vivo Protein Turnover Rates

Isotopic labeling with deuterium oxide (D2O) is a common technique for estimating in vivo protein turnover, but its use has been limited by two long-standing problems: (1) identifying non-monoisotopic peptides; and (2) estimating protein turnover rates in the presence of dynamic amino acid enrichment. In this paper, we present a novel experimental and analytical framework for solving these two problems. Peptides with high probabilities of labeling in many amino acids present fragmentation spectra that frequently do not match the theoretical spectra used in standard identification algorithms. We resolve this difficulty using a modified search algorithm we call Conditional Ion Distribution Search (CIDS). Increased identifications from CIDS along with direct measurement of amino acid enrichment and statistical modeling that accounts for heterogeneous information across peptides, dramatically improves the accuracy and precision of half-life estimates. We benchmark the approach in cells, where near-complete labeling is possible, and conduct an in vivo experiment revealing, for the first time, differences in protein turnover between mice and naked mole-rats commensurate with their disparate longevity.

bioinformatics↗