Bioinformatics & AI

Week of 6 May 2026

11 articles

Geno'X Veille — genox-veille.fr

Key takeaways this week

►AgentIA (Nature 2026) — agentic system for rare diseases — traceable step-by-step reasoning: resolution of complex cases with explainable audit trail; a milestone in clinical utility of diagnostic AI.
►AlphaGenome (Nature 2026) — unified prediction of regulatory variant effects on chromatin, splicing and expression: 10× more precise than previous models across 50+ functional modalities.
►PheMART (Nature Biomed Eng 2026) — phenotype-specific missense variant prediction via contrastive learning + EHR: outperforms AlphaMissense and CADD across 4,179 phenotypes; 5.1M variant database available.
►HiFi long-read RNA-seq EJHG 2026 — +8 diagnoses in 25 patients vs short-read RNA-seq: variant phasing on RNA, complete isoforms, intron retention — emerging gold standard for splicing variants.
►Franklin leads ACMG/AMP benchmark on 151 real clinical VCFs (Bioinformatics 2026): reassess InterVar if still used as sole classification tool.
►LLMs in genetics 2026 — Gemini 2.5 Pro resolves 6 rare disease cases without fine-tuning; Med-PaLM 2 discovers new genetic factor for deafness validated by knock-in mouse.
►New PVS1_RNA + BP7_S criteria (Clin Chem 2026) — ClinGen SVI Splicing Subgroup: SpliceAI prediction + RNA functional assay combination essential to reclassify BRCA1/BRCA2 splicing VUS.

Category:

Filters··Min. score

11 articles of 11

Sort

Rare diseases — diagnosis by agentic AI system

An agentic system for rare disease diagnosis with traceable reasoning

Rare diseases — diagnosis by agentic AI system

LLM appliedClinical pipeline

Nature, 2026· MarRead

Language model / AI

Advancing regulatory variant effect prediction with AlphaGenome

Regulatory variants — functional effect prediction on chromatin, splicing and expression

Language model / AIPathogenicity predictionNew tool

Nature, 2026· JanRead

Missense variants — phenotype-specific pathogenicity prediction

Phenotypic prediction of missense variants via deep contrastive learning

Missense variants — phenotype-specific pathogenicity prediction

Pathogenicity predictionNew tool

Nature Biomedical Engineering, 2026· AprRead

HiFi long-read RNA sequencing enhances clinical diagnostics in rare disorders

Rare diseases with suspected splicing variants — LRS vs SRS differential diagnostic yield

Long-readLong-read sequencingDiagnostic RNA-seq

European Journal of Human Genetics, 2026· MarRead

Population-level structural variant characterization using pangenome graphs

Structural variants (SV) — population-level catalog via pangenome

Long-readSV caller

Nature Genetics, 2026· MarRead

Rare pediatric genetic diseases — undiagnosed after WES/short-read WGS

Diagnostic utility of clinical genome reanalysis in rare pediatric disorders using long-read sequencing

Rare pediatric genetic diseases — undiagnosed after WES/short-read WGS

Long-read sequencingGenomic reanalysis

HGG Advances, 2026· AprRead

Clinical pipeline

Scaling genomic reanalysis to unlock diagnoses and transform rare disease care

Rare genetic diseases — optimization of diagnostic yield through systematic reanalysis

Clinical pipelineGenomic reanalysisDiagnostic yield

HGG Advances, 2026· AprRead

Targeted long-read RNA sequencing for rare disease diagnosis and variant interpretation

Rare genetic diseases — variant interpretation by targeted long-read RNA-seq

Long-readLong-read sequencingDiagnostic RNA-seq

Science Advances, 2026· AprRead

Variant interpretation

Comprehensive evaluation of ACMG/AMP-based variant classification tools

Mendelian diseases — ACMG/AMP classification tool benchmark in real clinical context

Variant interpretationACMG classificationBenchmark

Bioinformatics (Oxford), 2026· FebRead

Variant interpretation

Splicing Predictions, Splicing Assays, and Variant Classification Using ACMG/AMP Guidelines: Challenges Observed with BRCA1 and BRCA2 Variants

HBOC — reclassification of BRCA1/BRCA2 splicing variants

Variant interpretationACMG classificationVUS reclassified

Clinical Chemistry, 2026· FebRead

Language model / AI

Genetic Diagnosis and Discovery Enabled by Large Language Models

Rare diseases — LLM use for genetic diagnosis and new gene discovery

Language model / AILLM applied

References and sources

Rare diseases — diagnosis by agentic AI system. Nature, 2026. PMID 41708847. Score 9/10. https://pubmed.ncbi.nlm.nih.gov/41708847/
Regulatory variants — functional effect prediction on chromatin, splicing and expression. Nature, 2026. PMID 41606153. Score 9/10. https://pubmed.ncbi.nlm.nih.gov/41606153/
Missense variants — phenotype-specific pathogenicity prediction. Nature Biomedical Engineering, 2026. PMID 41981312. Score 9/10. https://pubmed.ncbi.nlm.nih.gov/41981312/
Rare diseases with suspected splicing variants — LRS vs SRS differential diagnostic yield. European Journal of Human Genetics, 2026. PMID 41807732. Score 9/10. https://pubmed.ncbi.nlm.nih.gov/41807732/
Structural variants (SV) — population-level catalog via pangenome. Nature Genetics, 2026. PMID 41807798. Score 8/10. https://pubmed.ncbi.nlm.nih.gov/41807798/
Rare pediatric genetic diseases — undiagnosed after WES/short-read WGS. HGG Advances, 2026. PMID 42050932. Score 8/10. https://pubmed.ncbi.nlm.nih.gov/42050932/
Rare genetic diseases — optimization of diagnostic yield through systematic reanalysis. HGG Advances, 2026. PMID 41715921. Score 8/10. https://pubmed.ncbi.nlm.nih.gov/41715921/
Mendelian diseases — ACMG/AMP classification tool benchmark in real clinical context. Bioinformatics (Oxford), 2026. PMID 41688871. Score 7/10. https://pubmed.ncbi.nlm.nih.gov/41688871/
HBOC — reclassification of BRCA1/BRCA2 splicing variants. Clinical Chemistry, 2026. PMID 41499258. Score 7/10. https://pubmed.ncbi.nlm.nih.gov/41499258/
Rare diseases — LLM use for genetic diagnosis and new gene discovery. 2026. PMID 41655254. Score 7/10. https://pubmed.ncbi.nlm.nih.gov/41655254/
Rare genetic diseases — variant interpretation by targeted long-read RNA-seq. Science Advances, 2026. PMID 41984969. Score 8/10. https://pubmed.ncbi.nlm.nih.gov/41984969/