#4 · Genomics

Internal ground truths benchmark RNA-seq transcript and expression inference tools

The dominant thread is benchmarking of computational tools that reconstruct transcript structure or quantify expression from RNA-seq data. TUSCO evaluates long-read transcriptome reconstruction using a curated set of single-isoform genes as endogenous ground truth. CATS assesses transcriptome assembly quality through reference-free and reference-based pipelines, benchmarked on 1056 simulated and public RNA-seq datasets. CrossFilt addresses a distinct error source by retaining only reciprocally mapping reads across primate genomes, reducing alignment bias in cross-species expression inference. What is new is the reliance on internal reference standards and multi-pipeline assessment to judge quantification and assembly accuracy. Pan1C and Verticall address pangenome graphs and bacterial recombination, respectively, and do not share this focus.