Biology subjects

Ahsan, M. U.

Publications and source records attributed to Ahsan, M. U..

4 recordsLinked to original sources

bioRxiv2026

A protein language model for domain, functional site, and variant effect prediction

Protein sequence databases are growing far faster than experimentally curated annotations, limiting functional interpretation of newly identified proteins and coding variants. We developed PALM, a protein annotation language model that jointly predicts domain and functional-site occurrence, while preserving overlapping and nested annotations through a soft-target objective. PALM integrates ESM2 with low-rank adaptation (LoRA), learned layer aggregation, and a bidirectional long short-term memory (BiLSTM) encoder. PALM was trained on ~15 million InterPro-annotated UniRef50 sequences spanning 17,951 domain and 1,383 site classes. Across 2.3 million held-out proteins, PALM achieved domain- and site-occurrence F1 scores of 0.941 and 0.877, respectively, and retained substantial performance in proteins with <20% sequence identity to the training and validation data. PALM also substantially outperformed ESM3 annotation baseline. Without training on variant-pathogenicity labels, PALM embedding disruption distinguished 151,823 ClinVar missense variants more accurately than ESM2 embedding disruption (AUROC 0.835 versus 0.710), and combining PALM embedding disruption with the ESM2 likelihood-ratio signal reached an AUROC of 0.877. The same combined zero-shot score outperformed AlphaMissense on a benchmark dataset on 128,161 missense variants, that were derived from literature but not present in ClinVar database. At the protein level, PALM embeddings showed stronger CATH-Gene3D superfamily organization than frozen ESM2 and ESM3 embeddings, and improved prediction of experimental activation and repression effects when combined with ESM2. In a downstream case study, we demonstrated the ability to use PALM-derived embedding as proxy of deep mutational scanning (DMS), showcasing the broad application of PALM. Together, PALM converts curated residue-level knowledge into transferable and mechanistically interpretable representations that connect protein sequence, function, and variant consequence.

bioinformatics↗

bioRxiv2026

Long-Read Transcriptome Sequencing and Functional Validation Reveals Novel and Oncogenic Gene Fusions in Fusion Panel-Negative Gliomas

Gliomas comprise a heterogeneous group of central nervous system tumors in which gene fusions (GFs) are significant oncogenic drivers and emerging diagnostic and therapeutic biomarkers. In cancer diagnosis, GF detection largely relies on targeted short-read sequencing fusion panels, such as the Childrens Hospital of Philadelphia (CHOP) Fusion Panel (FUSIP). While these panels are effective for detecting recurrent, well-characterized GFs, they are limited to predefined gene sets and cannot identify full-length transcripts. Here, we analyzed 49 high-and low-grade gliomas previously classified as fusion-negative by FUSIP using an untargeted whole-transcriptome RNA sequencing approach with Oxford Nanopore Technologies (ONT) long-read sequencing. This enabled transcriptome-wide fusion discovery of additional known and potentially novel oncogenic GFs beyond panel constraints. Long-read sequencing further allowed direct resolution of full-length fusion transcripts and their associated isoform structures. By integrating GF detection with isoform-level transcript analysis, we identified fusion-associated transcript isoforms with alternative splicing patterns that aligned near reported GF breakpoints, including ZNF254::GNAS and PTPRK::NOX3, which have not been reported in literature or existing fusion databases. To assess functional relevance, candidate GFs were evaluated using the Drosophila melanogaster model, with ventral nerve cord (VNC) morphology serving as a quantitative in vivo readout of fusion-induced disruption of glial regulation. VNC enlargement or elongation reflects abnormal glial growth or defects in brain tissue organization. Of the 15 candidate GFs subjected to experimental functional testing, 8 induced significant VNC abnormalities relative to wild-type controls, indicating fusion-specific disruption and oncogenic potential. Notably, CLDND1::WRN and DUSP22::APOE produced the most pronounced VNC phenotypes. Together, these findings demonstrate that untargeted transcriptome-wide GF discovery, coupled with long-read isoform-level analysis and in vivo functional validation, enables the identification and prioritization of potentially novel and clinically relevant GFs that are missed by standard targeted short-read fusion panels in glioma.

cancer biology↗

bioRxiv2025

A Comparative Evaluation of Computational Models for RNA modification detection using Nanopore sequencing with RNA004 Chemistry

Direct RNA sequencing from Oxford Nanopore Technologies (ONT) has become a valuable method for studying RNA modifications such as N6-methyladenosine (m6A), pseudouridine ({psi}), and 5-methylcytosine (m5C). Recent advancements in the RNA004 chemistry substantially reduce sequencing errors compared to previous chemistries (e.g., RNA002), thereby promising enhanced accuracy for epitranscriptomic analysis. In this study, we benchmark the performance of two state-of-the-art RNA modification detection models capable of handling RNA004 data - ONTs Dorado and m6Anet - using two wild-type (WT) cell lines, HEK293T and HeLa, with respective ground truths from GLORI and eTAM-seq, and their paired in vitro transcribed (IVT) RNA as negative controls. We found that under default settings and considering sites with [≥]10% modification ratio and [≥]10X coverage, Dorado has higher recall ([~]0.92) than m6Anet ([~]0.51) for m6A detection. Among the overlapping methylated sites between ground truth and computational predictions, there are high correlations of site-specific m6A modification stoichiometry, with correlation coefficient of [~]0.89 for Dorado-truth comparison and [~]0.72 for m6Anet-truth comparison. However, combined assessment of WT and IVT datasets show that while the per-site false positive rate (FPR) can be lower ([~]8% for Dorado and [~]33% for m6Anet), both computational tools can have high per-site false discovery rate (FDR) of m6A ([~]40% for Dorado and [~]80% for m6Anet) due to the low prevalence of m6A in transcriptome, with a similar trend observed for pseudouridine ([~]95% FDR for Dorado). Additional motif analysis reveals that both Dorado and m6Anet exhibit high heterogeneity of false positive calls across sequence contexts, suggesting that sequence contexts help determine accuracy of specific modification calls. There is also a substantial overlap of false positive calls between the two IVT samples, suggesting a post-filtering strategy to improve modification calling by compiling a set of low-confidence sites with a probabilistic model from several IVT samples across diverse cells/tissues. Our analysis highlights key strengths and limitations of the current generation of m6A detection algorithms and offers insights into optimizing thresholds and interpretability. The IVT datasets generated by the RNA004 chemistry provides a publicly available benchmark resource for further development and refinement of computational methods.

bioinformatics↗

bioRxiv2024

LongReadSum: A fast and flexible quality control and signal summarization tool for long-read sequencing data

While several well-established quality control (QC) tools are available for short reads sequencing data, there is a general paucity of computational tools that provide long read metrics in a fast and comprehensive manner across all major sequencing platforms (such as PacBio, Oxford Nanopore, Illumina Complete Long Read) and data formats (such as ONT POD5, FAST5, basecall summary files and PacBio unaligned BAM). Additionally, none of the current tools provide support for summarizing Oxford Nanopore basecall signal or comprehensive base modification (methylation) information from genomic data. Furthermore, nowadays a single PromethION flowcell on the Oxford Nanopore platform can generate terabytes of signal data, which cannot be handled by existing tools designed for small-scale flowcells. To address these challenges, here we present LongReadSum, a multi-threaded C++ tool which provides fast and comprehensive QC reports on all major aspects of sequencing data (such as read, base, base quality, alignment, and base modification metrics) and produce basecalling signal intensity information from the Oxford Nanopore platform. We demonstrate use cases to analyze cDNA sequencing, direct mRNA sequencing, reduced representation methylation sequencing (RRMS) through adaptive sequencing, as well as whole genome sequencing (WGS) data using diverse long-read platforms.

bioinformatics↗