Accurate isoform quantification in bulk and single-cell RNA-seq by integrating multi-platform sequencing data and sequence-to-expression models

Abstract

Accurate quantification of transcript isoforms is crucial for understanding gene regulation, functional diversity, and cellular behavior. Existing RNA sequencing methods have significant limitations: short-read (SR) sequencing provides high depth but struggles with isoform deconvolution, especially in single-cell data with strong positional biases, whereas long-read (LR) sequencing offers isoform resolution at the cost of lower depth and platform-specific biases. Here, we introduce Multi-Platform Aggregation and Quantification of Transcripts (MPAQT), a probabilistic generative framework that integrates sequencing data across platforms while explicitly modeling their distinct data-generating processes and technical biases. Across extensive simulations and experimental benchmarks, MPAQT provides state-of-the-art transcript quantification from short-read data alone and further improves accuracy by jointly analyzing short- and long-read sequencing data in both bulk and single-cell transcriptomes. We show that long-read sequencing is particularly informative for resolving isoforms of genes with complex transcript structures, while integration with short reads mitigates biases and transcript dropouts inherent to long-read data. MPAQT can additionally incorporate sequence-to-expression models as informative priors that are iteratively refined by sequencing observations. Using models of transcript abundance based on RNA-binding protein recognition sequences and transcript co-expression patterns, we substantially improve isoform quantification from single-cell data and reveal a strong contribution of isoform-specific sequence features in untranslated regions to isoform abundance and exon usage. Together, our results show that complementary sequencing technologies and predictive models of transcript regulation jointly constrain latent transcript abundances, enabling accurate isoform-resolved quantification across bulk and single-cell RNA sequencing.