Length biases in single-cell RNA sequencing of pre-mRNA

Single-molecule pre-mRNA and mRNA sequencing data can be modeled and analyzed using the Markov chain formalism to yield genome-wide insights into transcription. However, quantitative inference with such data requires careful assessment and understanding of noise sources. We find that long pre-mRNA transcripts are over-represented in sequencing data, and explore the mechanistic implications. A biological explanation for this phenomenon within our modeling framework requires unrealistic transcriptional parameters, leading us to posit a length-based model of capture bias. We provide solutions for this model, and use them to find concordant and mechanistically plausible parameter trends across data from multiple single-cell RNA-seq experiments in several species.