American journal of human genetics最新文献

{"title":"A novel multivariable Mendelian randomization framework to disentangle highly correlated exposures with application to metabolomics.","authors":"Lap Sum Chan, Mykhaylo M Malakhov, Wei Pan","doi":"10.1016/j.ajhg.2024.07.007","DOIUrl":"10.1016/j.ajhg.2024.07.007","url":null,"abstract":"<p><p>Mendelian randomization (MR) utilizes genome-wide association study (GWAS) summary data to infer causal relationships between exposures and outcomes, offering a valuable tool for identifying disease risk factors. Multivariable MR (MVMR) estimates the direct effects of multiple exposures on an outcome. This study tackles the issue of highly correlated exposures commonly observed in metabolomic data, a situation where existing MVMR methods often face reduced statistical power due to multicollinearity. We propose a robust extension of the MVMR framework that leverages constrained maximum likelihood (cML) and employs a Bayesian approach for identifying independent clusters of exposure signals. Applying our method to the UK Biobank metabolomic data for the largest Alzheimer disease (AD) cohort through a two-sample MR approach, we identified two independent signal clusters for AD: glutamine and lipids, with posterior inclusion probabilities (PIPs) of 95.0% and 81.5%, respectively. Our findings corroborate the hypothesized roles of glutamate and lipids in AD, providing quantitative support for their potential involvement.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":" ","pages":"1834-1847"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11393695/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141896544","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Inflation of polygenic risk scores caused by sample overlap and relatedness: Examples of a major risk of bias.","authors":"Colin A Ellis, Karen L Oliver, Rebekah V Harris, Ruth Ottman, Ingrid E Scheffer, Heather C Mefford, Michael P Epstein, Samuel F Berkovic, Melanie Bahlo","doi":"10.1016/j.ajhg.2024.07.014","DOIUrl":"10.1016/j.ajhg.2024.07.014","url":null,"abstract":"<p><p>Polygenic risk scores (PRSs) are an important tool for understanding the role of common genetic variants in human disease. Standard best practices recommend that PRSs be analyzed in cohorts that are independent of the genome-wide association study (GWAS) used to derive the scores without sample overlap or relatedness between the two cohorts. However, identifying sample overlap and relatedness can be challenging in an era of GWASs performed by large biobanks and international research consortia. Although most genomics researchers are aware of best practices and theoretical concerns about sample overlap and relatedness between GWAS and PRS cohorts, the prevailing assumption is that the risk of bias is small for very large GWASs. Here, we present two real-world examples demonstrating that sample overlap and relatedness is not a minor or theoretical concern but an important potential source of bias in PRS studies. Using a recently developed statistical adjustment tool, we found that excluding overlapping and related samples was equal to or more powerful than adjusting for overlap bias. Our goal is to make genomics researchers aware of the magnitude of risk of bias from sample overlap and relatedness and to highlight the need for mitigation tools, including independent validation cohorts in PRS studies, continued development of statistical adjustment methods, and tools for researchers to test their cohorts for overlap and relatedness with GWAS cohorts without sharing individual-level data.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":" ","pages":"1805-1809"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11393675/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142016069","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Genetic modifiers of body mass index in individuals with cystic fibrosis.","authors":"Hua Ling,Karen S Raraigh,Elizabeth W Pugh,Melis A Aksit,Peng Zhang,Rhonda G Pace,Anna V Faino,Michael J Bamshad,Ronald L Gibson,Wanda O'Neal,Michael R Knowles,Scott M Blackman,Garry R Cutting,","doi":"10.1016/j.ajhg.2024.08.004","DOIUrl":"https://doi.org/10.1016/j.ajhg.2024.08.004","url":null,"abstract":"To identify modifier loci underlying variation in body mass index (BMI) in persons with cystic fibrosis (pwCF), we performed a genome-wide association study (GWAS). Utilizing longitudinal height and weight data, along with demographic information and covariates from 4,393 pwCF, we calculated AvgBMIz representing the average of per-quarter BMI Z scores. The GWAS incorporated 9.8M single nucleotide polymorphisms (SNPs) with a minor allele frequency (MAF) > 0.005 extracted from whole-genome sequencing (WGS) of each study subject. We observed genome-wide significant association with a variant in FTO (FaT mass and Obesity-associated gene; rs28567725; p value = 1.21e-08; MAF = 0.41, β = 0.106; n = 4,393 individuals) and a variant within ADAMTS5 (A Disintegrin And Metalloproteinase with ThromboSpondin motifs 5; rs162500; p value = 2.11e-10; MAF = 0.005, β = -0.768; n = 4,085 pancreatic-insufficient individuals). Notably, BMI-associated variants in ADAMTS5 occur on a haplotype that is much more common in African (AFR, MAF = 0.183) than European (EUR, MAF = 0.006) populations (1000 Genomes project). A polygenic risk score (PRS) calculated using 924 SNPs (excluding 17 in FTO) showed significant association with AvgBMIz (p value = 2.2e-16; r2 = 0.03). Association between variants in FTO and the PRS correlation reveals similarities in the genetic architecture of BMI in CF and the general population. Inclusion of Black individuals in whom the single-gene disorder CF is much less common but genomic diversity is greater facilitated detection of association with variants that are in LD with functional SNPs in ADAMTS5. Our results illustrate the importance of population diversity, particularly when attempting to identify variants that manifest only under certain physiologic conditions.","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":"7 1","pages":""},"PeriodicalIF":9.8,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142174661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Liver eQTL meta-analysis illuminates potential molecular mechanisms of cardiometabolic traits.","authors":"K Alaine Broadaway, Sarah M Brotman, Jonathan D Rosen, Kevin W Currin, Abdalla A Alkhawaja, Amy S Etheridge, Fred Wright, Paul Gallins, Dereje Jima, Yi-Hui Zhou, Michael I Love, Federico Innocenti, Karen L Mohlke","doi":"10.1016/j.ajhg.2024.07.017","DOIUrl":"10.1016/j.ajhg.2024.07.017","url":null,"abstract":"<p><p>Understanding the molecular mechanisms of complex traits is essential for developing targeted interventions. We analyzed liver expression quantitative-trait locus (eQTL) meta-analysis data on 1,183 participants to identify conditionally distinct signals. We found 9,013 eQTL signals for 6,564 genes; 23% of eGenes had two signals, and 6% had three or more signals. We then integrated the eQTL results with data from 29 cardiometabolic genome-wide association study (GWAS) traits and identified 1,582 GWAS-eQTL colocalizations for 747 eGenes. Non-primary eQTL signals accounted for 17% of all colocalizations. Isolating signals by conditional analysis prior to coloc resulted in 37% more colocalizations than using marginal eQTL and GWAS data, highlighting the importance of signal isolation. Isolating signals also led to stronger evidence of colocalization: among 343 eQTL-GWAS signal pairs in multi-signal regions, analyses that isolated the signals of interest resulted in higher posterior probability of colocalization for 41% of tests. Leveraging allelic heterogeneity, we predicted causal effects of gene expression on liver traits for four genes. To predict functional variants and regulatory elements, we colocalized eQTL with liver chromatin accessibility QTL (caQTL) and found 391 colocalizations, including 73 with non-primary eQTL signals and 60 eQTL signals that colocalized with both a caQTL and a GWAS signal. Finally, we used publicly available massively parallel reporter assays in HepG2 to highlight 14 eQTL signals that include at least one expression-modulating variant. This multi-faceted approach to unraveling the genetic underpinnings of liver-related traits could lead to therapeutic development.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":" ","pages":"1899-1913"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11393674/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142034892","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Calibration of variant effect predictors on genome-wide data masks heterogeneous performance across genes.","authors":"Malvika Tejura, Shawn Fayer, Abbye E McEwen, Jake Flynn, Lea M Starita, Douglas M Fowler","doi":"10.1016/j.ajhg.2024.07.018","DOIUrl":"10.1016/j.ajhg.2024.07.018","url":null,"abstract":"<p><p>In silico variant effect predictions are available for nearly all missense variants but played a minimal role in clinical variant classification because they were deemed to provide only supporting evidence. Recently, the ClinGen Sequence Variant Interpretation (SVI) Working Group updated recommendations for variant effect prediction use. By analyzing control pathogenic and benign variants across all genes, they were able to compute evidence strength for predictor score intervals with some intervals generating moderate, strong, or even very strong evidence. However, this genome-wide approach could obscure heterogeneous predictor performance in different genes. We quantified the gene-by-gene performance of two top predictors, REVEL and BayesDel, by analyzing control variants in each predictor score interval in 3,668 disease-relevant genes. Approximately 10% of intervals had sufficient control variants for analysis, and ∼70% of these intervals exceeded the maximum number of incorrect predictions implied by the SVI recommendations. These trending discordant intervals arose owing to the divergence of the gene-specific distribution of predictions from the genome-wide distribution, suggesting that gene-specific calibration is needed in many cases. Approximately 22% of ClinVar missense variants of uncertain significance in genes we analyzed (REVEL = 100,629, BayesDel = 71,928) had predictions in trending discordant intervals. Thus, genome-wide calibrations could result in many variants receiving inappropriate evidence strength. To facilitate a review of the SVI's calibrations, we developed a web application enabling visualization of gene-specific predictions and trending concordant and discordant intervals.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":" ","pages":"2031-2043"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11393694/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142034891","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evidence-based recommendations for gene-specific ACMG/AMP variant classification from the ClinGen ENIGMA BRCA1 and BRCA2 Variant Curation Expert Panel.","authors":"Michael T Parsons, Miguel de la Hoya, Marcy E Richardson, Emma Tudini, Michael Anderson, Windy Berkofsky-Fessler, Sandrine M Caputo, Raymond C Chan, Melissa S Cline, Bing-Jian Feng, Cristina Fortuno, Encarna Gomez-Garcia, Johanna Hadler, Susan Hiraki, Megan Holdren, Claude Houdayer, Kathleen Hruska, Paul James, Rachid Karam, Huei San Leong, Alexandra Martins, Arjen R Mensenkamp, Alvaro N Monteiro, Vaishnavi Nathan, Robert O'Connor, Inge Sokilde Pedersen, Tina Pesaran, Paolo Radice, Gunnar Schmidt, Melissa Southey, Sean Tavtigian, Bryony A Thompson, Amanda E Toland, Clare Turnbull, Maartje J Vogel, Jamie Weyandt, George A R Wiggins, Lauren Zec, Fergus J Couch, Logan C Walker, Maaike P G Vreeswijk, David E Goldgar, Amanda B Spurdle","doi":"10.1016/j.ajhg.2024.07.013","DOIUrl":"10.1016/j.ajhg.2024.07.013","url":null,"abstract":"<p><p>The ENIGMA research consortium develops and applies methods to determine clinical significance of variants in hereditary breast and ovarian cancer genes. An ENIGMA BRCA1/2 classification sub-group, formed in 2015 as a ClinGen external expert panel, evolved into a ClinGen internal Variant Curation Expert Panel (VCEP) to align with Food and Drug Administration recognized processes for ClinVar contributions. The VCEP reviewed American College of Medical Genetics and Genomics/Association of Molecular Pathology (ACMG/AMP) classification criteria for relevance to interpreting BRCA1 and BRCA2 variants. Statistical methods were used to calibrate evidence strength for different data types. Pilot specifications were tested on 40 variants and documentation revised for clarity and ease of use. The original criterion descriptions for 13 evidence codes were considered non-applicable or overlapping with other criteria. Scenario of use was extended or re-purposed for eight codes. Extensive analysis and/or data review informed specification descriptions and weights for all codes. Specifications were applied to pilot variants with pre-existing ClinVar classification as follows: 13 uncertain significance or conflicting, 14 pathogenic and/or likely pathogenic, and 13 benign and/or likely benign. Review resolved classification for 11/13 uncertain significance or conflicting variants and retained or improved confidence in classification for the remaining variants. Alignment of pre-existing ENIGMA research classification processes with ACMG/AMP classification guidelines highlighted several gaps in the research processes and the baseline ACMG/AMP criteria. Calibration of evidence strength was key to justify utility and strength of different data types for gene-specific application. The gene-specific criteria demonstrated value for improving ACMG/AMP-aligned classification of BRCA1 and BRCA2 variants.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":" ","pages":"2044-2058"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11393667/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141981497","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Implementation of a dyadic nomenclature for monogenic diseases.","authors":"Courtney Thaxton, Leslie G Biesecker, Marina DiStefano, Melissa Haendel, Ada Hamosh, Emma Owens, Sharon E Plon, Heidi L Rehm, Jonathan S Berg","doi":"10.1016/j.ajhg.2024.07.019","DOIUrl":"10.1016/j.ajhg.2024.07.019","url":null,"abstract":"<p><p>A core task when establishing the strength of evidence for a gene's role in a monogenic disorder is determining the appropriate disease entity to curate. Establishing this concept determines which evidence can be applied and quantified toward the final gene-disease validity, variant pathogenicity, or actionability classification. Genes with implications in more than one phenotype can necessitate a process of lumping and splitting, disease reorganization, and updates to disease nomenclature. Reappraisal of the names that are used as labels for disease entities is therefore a necessary and perpetual process. The Clinical Genome Resource (ClinGen), in collaboration with representatives from Monarch Disease Ontology (Mondo) and Online Inheritance in Man (OMIM), formed the Disease Naming Advisory Committee (DNAC) to develop guidance for groups faced with the need to establish the \"curated disease entity\" for gene-phenotype validity and variant pathogenicity and to update disease names for clinical use when necessary. The objective of this group was to harmonize guidance for disease naming across these nosologic entities and among ClinGen curation groups in collaboration with other disease-related professional groups. Here, we present the initial guidance developed by the DNAC with representative examples provided by the ClinGen expert panels and working groups that warranted nomenclature updates. We also discuss the broader implications of these efforts and their benefits for harmonization of gene-disease validity curation. Overall, this work sheds light on current inconsistencies and/or discrepancies and is designed to engage the broader community on how ClinGen defines monogenic disorders using a consistent approach for disease naming.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":"111 9","pages":"1810-1818"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11393707/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142144959","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SMAD4 mutations causing Myhre syndrome are under positive selection in the male germline.","authors":"Katherine A Wood, R Spencer Tong, Marialetizia Motta, Viviana Cordeddu, Eleanor R Scimone, Stephen J Bush, Dale W Maxwell, Eleni Giannoulatou, Viviana Caputo, Alice Traversa, Cecilia Mancini, Giovanni B Ferrero, Francesco Benedicenti, Paola Grammatico, Daniela Melis, Katharina Steindl, Nicola Brunetti-Pierri, Eva Trevisson, Andrew Om Wilkie, Angela E Lin, Valerie Cormier-Daire, Stephen Rf Twigg, Marco Tartaglia, Anne Goriely","doi":"10.1016/j.ajhg.2024.07.006","DOIUrl":"10.1016/j.ajhg.2024.07.006","url":null,"abstract":"<p><p>While it is widely thought that de novo mutations (DNMs) occur randomly, we previously showed that some DNMs are enriched because they are positively selected in the testes of aging men. These \"selfish\" mutations cause disorders with a shared presentation of features, including exclusive paternal origin, significant increase of the father's age, and high apparent germline mutation rate. To date, all known selfish mutations cluster within the components of the RTK-RAS-MAPK signaling pathway, a critical modulator of testicular homeostasis. Here, we demonstrate the selfish nature of the SMAD4 DNMs causing Myhre syndrome (MYHRS). By analyzing 16 informative trios, we show that MYHRS-causing DNMs originated on the paternally derived allele in all cases. We document a statistically significant epidemiological paternal age effect of 6.3 years excess for fathers of MYHRS probands. We developed an ultra-sensitive assay to quantify spontaneous MYHRS-causing SMAD4 variants in sperm and show that pathogenic variants at codon 500 are found at elevated level in sperm of most men and exhibit a strong positive correlation with donor's age, indicative of a high apparent germline mutation rate. Finally, we performed in vitro assays to validate the peculiar functional behavior of the clonally selected DNMs and explored the basis of the pathophysiology of the different SMAD4 sperm-enriched variants. Taken together, these data provide compelling evidence that SMAD4, a gene operating outside the canonical RAS-MAPK signaling pathway, is associated with selfish spermatogonial selection and raises the possibility that other genes/pathways are under positive selection in the aging human testis.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":" ","pages":"1953-1969"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11444041/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141905611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Genetic effects on the skin methylome in healthy older twins.","authors":"Christopher J Shore, Sergio Villicaña, Julia S El-Sayed Moustafa, Amy L Roberts, David A Gunn, Veronique Bataille, Panos Deloukas, Tim D Spector, Kerrin S Small, Jordana T Bell","doi":"10.1016/j.ajhg.2024.07.010","DOIUrl":"10.1016/j.ajhg.2024.07.010","url":null,"abstract":"<p><p>Whole-skin DNA methylation variation has been implicated in several diseases, including melanoma, but its genetic basis has not yet been fully characterized. Using bulk skin tissue samples from 414 healthy female UK twins, we performed twin-based heritability and methylation quantitative trait loci (meQTL) analyses for >400,000 DNA methylation sites. We find that the human skin DNA methylome is on average less heritable than previously estimated in blood and other tissues (mean heritability: 10.02%). meQTL analysis identified local genetic effects influencing DNA methylation at 18.8% (76,442) of tested CpG sites, as well as 1,775 CpG sites associated with at least one distal genetic variant. As a functional follow-up, we performed skin expression QTL (eQTL) analyses in a partially overlapping sample of 604 female twins. Colocalization analysis identified over 3,500 shared genetic effects affecting thousands of CpG sites (10,067) and genes (4,475). Mediation analysis of putative colocalized gene-CpG pairs identified 114 genes with evidence for eQTL effects being mediated by DNA methylation in skin, including in genes implicating skin disease such as ALOX12 and CSPG4. We further explored the relevance of skin meQTLs to skin disease and found that skin meQTLs and CpGs under genetic influence were enriched for multiple skin-related genome-wide and epigenome-wide association signals, including for melanoma and psoriasis. Our findings give insights into the regulatory landscape of epigenomic variation in skin.</p>","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":" ","pages":"1932-1952"},"PeriodicalIF":8.1,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11393713/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141974864","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessing the utility of large language models for phenotype-driven gene prioritization in the diagnosis of rare genetic disease.","authors":"Junyoung Kim,Kai Wang,Chunhua Weng,Cong Liu","doi":"10.1016/j.ajhg.2024.08.010","DOIUrl":"https://doi.org/10.1016/j.ajhg.2024.08.010","url":null,"abstract":"Phenotype-driven gene prioritization is fundamental to diagnosing rare genetic disorders. While traditional approaches rely on curated knowledge graphs with phenotype-gene relations, recent advancements in large language models (LLMs) promise a streamlined text-to-gene solution. In this study, we evaluated five LLMs, including two generative pre-trained transformers (GPT) series and three Llama2 series, assessing their performance across task completeness, gene prediction accuracy, and adherence to required output structures. We conducted experiments, exploring various combinations of models, prompts, phenotypic input types, and task difficulty levels. Our findings revealed that the best-performed LLM, GPT-4, achieved an average accuracy of 17.0% in identifying diagnosed genes within the top 50 predictions, which still falls behind traditional tools. However, accuracy increased with the model size. Consistent results were observed over time, as shown in the dataset curated after 2023. Advanced techniques such as retrieval-augmented generation (RAG) and few-shot learning did not improve the accuracy. Sophisticated prompts were more likely to enhance task completeness, especially in smaller models. Conversely, complicated prompts tended to decrease output structure compliance rate. LLMs also achieved better-than-random prediction accuracy with free-text input, though performance was slightly lower than with standardized concept input. Bias analysis showed that highly cited genes, such as BRCA1, TP53, and PTEN, are more likely to be predicted. Our study provides valuable insights into integrating LLMs with genomic analysis, contributing to the ongoing discussion on their utilization in clinical workflows.","PeriodicalId":7659,"journal":{"name":"American journal of human genetics","volume":"31 1","pages":""},"PeriodicalIF":9.8,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142170697","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}