Khan, I. M. | Search bioRxiv

bioRxiv2025

Boosting Transcript Assembly via Delineating Transcript Start and End Sites

Transcript assembly remains a challenging task despite the development of numerous methods. A major contributor to low assembly accuracy is the difficulty in accurately determining transcript start sites (TSSs) and end sites (TESs), due to the weak and noisy signals typically found in RNA-seq data. We present Telos, a two-stage machine learning framework for precise detection of TSSs and TESs and for transcript ranking. The method takes as input any assembly, typically generated by an existing assembler. In the first stage, Telos scores the TSSs and TESs in the input assembly using a machine learning model trained on a rich set of engineered features. These site-level scores will be passed to the second stage for transcript-level evaluation. In its second stage, Telos scores the entire transcripts by training another model that integrates features of their TSS and TES (including the inferred probabilities from the first stage), along with transcript abundance estimated by the assembler and statistics about exon lengths. We extensively evaluated Telos on ONT (cDNA and direct RNA), PacBio, and Illumina short-read RNA-seq datasets. In all cases, it consistently outperformed baseline methods. Telos is agile, but achieves substantial improvements, demonstrating the value of explicitly modeling TSS and TES, a gap in current transcript assembly tools. Telos can be paired with any assembler to accurately score the assembled transcripts. It is modular, easily extensible to emerging sequencing technologies, and hence we anticipate its broad adoption in transcriptomic studies.

bioinformatics↗

bioRxiv2024

Augmenting Transcriptome Annotations through the Lens of Splicing Evolution

Alternative splicing (AS) is a ubiquitous mechanism in eukaryotes. It is estimated that 90% of human genes are alternatively spliced. Despite enormous efforts, transcriptome annotations remain, nevertheless, incomplete. Conventional means of annotation were largely driven by experimental data such as RNA-seq and protein sequences, while little insight was shed on understanding transcriptomes and alternative splicings from the perspective of evolution. This study addresses this critical gap by presenting TENNIS (Transcript EvolutioN for New Isoform Splicing), an evolution-based model to predict unannotated isoforms and refine existing annotations without requiring additional data. The model of TENNIS is based on two minimal premises-AS isoforms evolve sequentially from existing isoforms, and each evolutionary step involves a single AS event. We formulate the identification of missing transcripts as an optimization problem and parsimoniously find the minimal number of novel transcripts. Our analysis showed approximately 80% of multi-transcript groups from six transcriptome annotations satisfy our evolutionary model. At a high confidence level, 40% of isoforms predicted by TENNIS were validated by deep long-read RNA-seq. In a simulated incomplete annotation scenario, TENNIS dramatically outperforms two randomized baseline approaches by a 2.25-3 fold-change in precision or a 3.5-3.9 fold-change in recall, after controlling the same level of recall or precision of the baseline methods. These results demonstrate that TENNIS effectively identifies missing transcripts by complying with minimal propositions, offering a powerful approach for transcriptome augmentations through the lens of alternative splicing evolutions. TENNIS is freely available at https://github.com/Shao-Group/tennis.

bioinformatics↗