This project was led by the talented @jakobschuster.bsky.social in colab with @qgouil.bsky.social & @mritchieau.bsky.social

As an example of the power of Matchbox, Jakob’s last tool was Restrander. A 17 line matchbox script achieved the same result as 1400 lines of C++ in Restrander!

We’ve demonstrated Matchbox for demultiplexing long-read scRNA-seq data with 10X or SPLiT-seq barcodes; restranding RNA-seq reads; assessing CRISPR editing efficiency; and haplotyping repeat regions.

Matchbox is implemented in rust and available from github.com/jakob-schust....

Sequencing data requires read processing. Some tasks are common (trimming, de-multiplexing, filtering) & others bespoke, especially if you have non-standard read-structures.

Introducing Matchbox: a fast and incredibly versatile read processor that can do all this & more 🧪
doi.org/10.1101/2025...

Research has shown genetic risk for disease can be imparted at the isoform level, as well as the gene level. Therefore, understanding which isoforms genes express is essential to correctly determining the disease-associated isoforms and the molecular mechanisms behind disease aetiology.

Overall we found more than 300 previously unreported RNA isoforms from 31 genes in brain. Some were highly abundant or even the dominant isoform, and we could show translation of novel RNAs into novel proteoforms, including new isoforms of the depression risk gene ITIH4 (see image).

- Different RNA isoforms from the same gene can have different modification rates at the same m6A site. Distance to a splice site, transcript 3' end, and the CDS versus UTR status of a nucleotide all exert influence.