Integrating RNA analysis with clinical exome sequencing could resolve over 5% of uncertain variants, significantly boosting diagnostic accuracy for rare disease patients. bit.ly/4lgjsH8 #GIMO #ExomeSequencing #RareDisease #RNA #VariantClassification

C. elegans model enables rapid reclassification of FH gene variants, offering a powerful tool to interpret #VUS and improve diagnosis in FH-associated metabolic diseases. bit.ly/3M9lpHT #GIMO #FHfum1 #ClinicalVariant #CRISPR #VariantClassification #ClinicalDiagnosis

Out of sight out of mind. Shortening the time from consent to #results return may improve uptake of #secondaryfindings return. bit.ly/4ftAqPn #variantclassification

This study evaluated GPT-4o, Llama 3.1, and Qwen 2.5 for classifying cancer genetic variants from OncoKB, CIViC, and a real-world #FoundationOne CDx dataset. @Nature_NPJ #Onco404 #Cancer #Kanser #PrecisionOncology #AI #MedTwitter #Genomics #LLM #VariantClassification #CDx

Variant Classification Using Proteomics-Informed Large Language Models Increases Power of Rare Variant Association Studies and Enhances Target Discovery 1. This study introduces a proteomics-informed refinement of large language models (LLMs) to improve the classification of rare missense variants and enhance the power of rare variant association studies in human genetics. 2. The authors use plasma proteomics data from 46,665 individuals in the UK Biobank to correlate protein abundance changes with coding variants, finding strong associations between predicted deleteriousness and proteomic readouts. 3. A two-step model was built: first, an ensemble classifier trained on synonymous and pLoF variants was used to label rare missense variants; second, these labels were used to fine-tune the ESM-1b LLM, producing the ESM-1b proteomics model. 4. ESM-1b proteomics achieved higher correlation with validation proteomic assay results than standard LLMs (ESM-1b, ESM-1v, AlphaMissense), particularly in within-gene analyses of missense variant impact. 5. When benchmarked on 241 gene-trait pairs with known pLoF associations, ESM-1b proteomics recapitulated 88 associations using only singleton missense variants—outperforming all tested methods including AlphaMissense (87) and ESM-1b (83). 6. Applied to 10 complex traits in the UK Biobank, the model yielded 177 gene-trait associations at genome-wide significance, a 24.6% increase over conventional ensemble methods and a 15.7% improvement over ESM-1b. 7. Novel associations identified by the model include PCSK6 with triglyceride levels and SIX1 with hearing loss—findings missed by conventional variant classifiers, highlighting its value in target discovery. 8. ESM-1b proteomics also outperformed standard ESM-1b in classifying ClinVar variants, achieving an AUROC of 0.940 vs 0.919. Though AlphaMissense had a slightly higher AUROC (0.947), ESM-1b p showed superior performance in association studies. 9. The approach generalized to other LLMs, including ESM-1v and ESM-2 models, with proteomic fine-tuning notably improving performance, especially in smaller models. 10. This work establishes that large-scale human proteomics can be a powerful, unbiased supervisory signal to refine LLMs for variant interpretation, improving both discovery yield and mechanistic insight in genetics. 📜Paper: biorxiv.org/content/10.1101/… #Genomics #Proteomics #VariantClassification #LLM #RareVariants #FunctionalGenomics #ProteinLanguageModels #HumanGenetics #UKBiobank #Bioinformatics #PrecisionMedicine #TargetDiscovery

To report or not to report? The utilization of #VUS subclasses & the development of subclass-specific professional guidance are crucial for improving patient diagnosis and resource utilization bit.ly/44eexA7 #variantclassification #reclassification @HeidiRehm

ENIGMA VCEP guidelines significantly improve BRCA1/BRCA2 variant interpretation, reducing VUS rates and streamlining clinical diagnostics bit.ly/4bHELwk #GIMO #ACMGAMP #BRCA1 #BRCA2 #VUS #ColdSpot #VariantClassification #UCSCGenomeBrowser #ClinicalGenetics

"The impact of systematized generation, evaluation, and incorporation of #machinelearning algorithms for #clinical #variantclassification" in #germline #genetictesting, #LaureFresard et al. Just one of the reasons @Invitae/#LabcorpGenetics/@Labcorp has the highest published variant classification accuracy for #germlinetesting, exceeding the standards set by @TheACMG & @AMPath (jamanetwork.com/journals/jam… @JAMANetworkOpen ) #access #reducedisparities #precisionmedicine #precisiontherapy #clinicaltrials #precisionprevention #AI #precisionclassification medrxiv.org/content/10.1101/…

"Reducing uncertainty in #variantclassification by leveraging the capabilities of #evidencemodeling" presented by @LaureFresard @Invitae @Labcorp @GeneticsSociety #ASHG24 eppro02.ativ.me/appinfo.php?… #access #reducedisparities #germlinetesting #genetictesting #precisionmedicine #AI #precisioninterpretation

Use of PP3/BP4 evidence from calibrated computational prediction tools has little impact on the number of variants classified as Pathogenic or Likely Pathogenic #ACMG/AMP #variantclassification #computational predictors bit.ly/3ZtV0cn

Excellent work by @CRowlandsBio and team at ICR and CanVIG - if you do variant classification, then this tool is a must #PS4 #casecontrolanalysis #variantclassification #CanVIG #cancer #cancergenetics @clare__turnbull @MirandaDurkie @DrAliceGarrett pubmed.ncbi.nlm.nih.gov/3922…

We’re thrilled to have Dr Leslie Biesecker from the @NIH present “Evolving guidelines for variant classification” at our next #DNAdialogue seminar! 📅 Thursday 30 May, 9am (AEST) Register here: tinyurl.com/5bxcha6u @genome_gov #genomics #SeminarSeries #variantclassification

ICYMI: Briana Marmelstein and Avi Anantharajah shared the collaboration that led to a patient's CDH1 variant being classified. Find out why it takes a village. #genetictesting #variantclassification #breastcancer #gastriccancer #RNA #DNA #Genomics hubs.ly/Q02nH3Br0

No tool can do it all. Thorough #biocuration still relies on multiple #literaturemining tools bit.ly/3OWIJai #RYR1 #variantclassification

Specific #epigenetic signature in CTCF-related autosomal dominant intellectual developmental disorder-21: another tool for diagnosis and #variantclassification bit.ly/47CaAmx

Recommendations from the #ClinGen Low Penetrance/ #RiskAllele working group support harmonized interpretation, #variantclassification, and reporting for low penetrance variants bit.ly/3ShSyBV #GIM

"Bringing multiplex assays of variant effect (#MAVE) data into #clinical #variantinterpretation" #DrJasonReuter @Invitae #ASHG23 @GeneticsSociety "...#machinelearning-based functional modeling platform quality control for MAVE datasets...incorporat[ing] them into #clinical #variantinterpretation... incorporat[ing] functional evidence for thousands of variants...to improve the #accuracy of #clinical #variantclassification. #patient #access #reducedisparities #germlinetesting #genetictesting #precisiondiagnosis #precisionmedicine #precisiontherapy #clinicaltrials #precisionprevention

Shin Hayashi shows novel approach to rapidly generate transgenic mouse models mimicking #VUS seen in patients to clarify pathogenicity. Outstanding work and approach to characterize tricky #VUS variants. #Cas9 #humangenetics #variantclassification #ASHG23 @GeneticsSociety

Hi #GCchat and other users of #ClinVar. You'll want to watch this video we just launched to train clinicians & researchers on the common sources of #GermLine #VariantClassification errors within the context of ClinVar entries. Please share it with others. youtube.com/watch?v=eHsWrS44…

Non-truncating variants and likely gene disrupting variants in #SMARCC2 produce two distinct phenotypes #Neurodevelopmental disorder #raredisease #variantclassification @the_elbo @berntpopp bit.ly/3r7yMhB