#11 · Genomics
Learning-based tools refine genome annotation, splicing, and variant calls
The common thread is computational annotation and filtering of genomic and transcriptomic features, with learned models increasingly applied to long-read data. TOGA2 and EviAnn address gene structure and orthology across vertebrate and eukaryotic genomes, while EDTA v2 extends transposable element annotation to animal genomes by adding LINE and SINE detection. On the RNA side, SpliSync uses the HyenaDNA genomic language model with a U-net to correct splice-site errors in long reads, and IRCAS performs reference-free detection and classification of alternative splicing. Deep learning post-call filters appear in VCboost for SNP and indel calls and Dual-SVF for structural variants, the latter fusing sequence content with alignment topology. DeepKOALA applies a GRU-based open-set framework to KEGG Orthology assignment, reporting F1 of 83.37% and a 36.5-fold speedup. The new elements are scale, reduced resource use, and combining sequence with alignment or evidence features; the items do not benchmark these tools against one another.
Published papers
Efficient evidence-based genome annotation with EviAnn.
EviAnn is an open-source evidence-based eukaryotic gene annotation system that builds exon-intron structures from transcript alignments or protein homology,…
IRCAS: a novel end-to-end approach to identify, rectify, and classify comprehensive alternative splicing events in a transcriptome without genome reference.
IRCAS is an end-to-end reference-free framework for alternative splicing detection, rectification, and classification in transcriptomes, using colored de…
VCboost: reducing false positives in long-read variant calling for single-nucleotide polymorphism and indel detection in challenging genomic regions.
VCboost, a deep learning post-calling filter, reduces false-positive SNP and indel calls from long-read sequencing, improving precision and F1 in challenging…
DeepKOALA: a scalable deep learning framework for KEGG Orthology assignment.
DeepKOALA: a GRU-based deep learning tool for KEGG Orthology assignment framed as open-set recognition, reporting F1 83.37% and 36.5-fold speedup over…
Dual-SVF: a robust knowledge-guided multimodal method for structural variant filtering in long-read sequencing.
Dual-SVF is an open-source, VCF-compatible multimodal method that filters long-read structural variant calls by fusing sequence content with alignment topology…
Preprints
Accurate, comprehensive gene annotation and ortholog identification across thousands of vertebrate genomes with TOGA2
TOGA2 is a scalable tool integrating exon-level gene annotation and orthology inference across thousands of vertebrate genomes, with reduced memory and runtime.
EDTA v2: enabling scalable TE annotation in animal genomes
EDTA v2 extends automated transposable element annotation to animal genomes by adding LINE/SINE detection, rewriting TIR-Learner, and speeding structural…
SpliSync: Genomic language model-driven splice site correction of long RNA sequencing reads
SpliSync corrects splice-site errors in long RNA reads using a genomic language model (HyenaDNA) with a U-net, improving alignment precision and alternative…
Topic history
| Date | Rank |
|---|---|
| #11 |