The brain's transcriptomic complexity remains poorly characterized at the isoform level: short reads (around 200 base pairs) cannot assemble full transcripts, while manual long-read sequencing protocols are labor-intensive and sensitive to RNA degradation in postmortem tissue. Work has already emerged around this challenge, from automating the analysis of cellular data to applying neural networks to physical signals, but no integrated, scalable solution existed for postmortem human brain.
SALRR automates Oxford Nanopore cDNA library preparation on the Hamilton Microlab NGS STAR platform, cutting hands-on time by 67% and raising throughput from 8 to 24 libraries per person per day. The Snakemake computational side includes Dorado basecalling, Pychopper filtering of full-length reads, minimap2 alignment to the GRCh38 genome, transcript reconstruction with IsoQuant and StringTie followed by merging via Isomatch, and classification of novel isoforms into structural categories with SQANTI. SIRV-Set 4 spike-in controls were used for quality control and calibration.
Across 13 libraries, an average of 68.8 million reads was obtained, with a median length of 2,064 bp, Q24.91 quality and a mapping rate of 99.93%. Automated preparation matched manual preparation in library yield and mapping rate, only slightly reducing the median length of aligned reads (1.7 versus 2.0 kb). At RIN values from 8.2 to 10, library yield correlated with RIN (r=0.70), but read counts and annotated transcript numbers remained stable. Expression correlation with short reads was r=0.742 at the gene level and r=0.760 at the transcript level; quantification accuracy against ERCC controls was r=0.939-0.943. A total of 31,607 expressed isoforms from 10,075 genes was detected, 8,532 of them novel (27%), dominated by intron retention, novel combinations of known junctions and alternative 3' ends. At the GBA1, CCNF, CHCHD10 and TREM2 loci, long reads resolved more isoforms than short reads, including two novel GBA1 isoforms (ISOMT_8486 and ISOMT_8487) with an alternative 5' structure.
Unlike neighboring work applying machine learning to cellular and physical data, SALRR tackles a purely biological problem: it makes long-read brain sequencing reproducible and scalable. It is the first platform to simultaneously address the low throughput of manual protocols, sensitivity to RNA quality and the fragmented nature of computational tools. It opens the door to population studies of isoform diversity in neurodegeneration that were previously unfeasible.
SALRR's modular architecture allows new technologies to be plugged in without redesign. Extending it to single-cell and spatial long-read RNA sequencing will resolve cell-type-specific isoform usage, which is critical for understanding the heterogeneity of neurodegenerative diseases. The platform is already being applied to hundreds of brain samples from Alzheimer's disease and related dementias through the CARD Long-Read Sequencing Initiative.
The results will influence neurogenomics, population transcriptomics and neurodegenerative disease research by providing isoform resolution that short reads cannot achieve.
Apply SALRR to large postmortem brain cohorts, including hundreds of CARD samples, and integrate the data with matched long-read genomic and epigenomic profiles from the same donors.
The work speaks directly to an unsolved problem in neurogenomics: how isoform switching affects neuronal function and disease predisposition when overall gene expression stays the same. SALRR provides a tool for systematically mapping such events in the human brain.
The median poly(A) tail length of nuclear transcripts was 143 nucleotides, while that of mitochondrial transcripts was just 41, reflecting a known feature of mitochondrial mRNAs.
In Greg Egan's novel Permutation City, copying a mind requires fully decoding the molecular states of neurons; SALRR brings reality closer to that level of detail by revealing isoform diversity that was previously invisible.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.