Artificial intelligence enables comprehensive genome interpretation and nomination of candidate diagnoses for rare genetic diseases.

Clinical interpretation of genetic variants in the context of the patient's phenotype is becoming the largest component of cost and time expenditure for genome-based diagnosis of rare genetic diseases. Artificial intelligence (AI) holds promise to greatly simplify and speed genome interpretation by integrating predictive methods with the growing knowledge of genetic disease. Here we assess the diagnostic performance of Fabric GEM, a new, AI-based, clinical decision support tool for expediting genome interpretation. We benchmarked GEM in a retrospective cohort of 119 probands, mostly NICU infants, diagnosed with rare genetic diseases, who received whole-genome or whole-exome sequencing (WGS, WES). We replicated our analyses in a separate cohort of 60 cases collected from five academic medical centers. For comparison, we also analyzed these cases with current state-of-the-art variant prioritization tools. Included in the comparisons were trio, duo, and singleton cases. Variants underpinning diagnoses spanned diverse modes of inheritance and types, including structural variants (SVs). Patient phenotypes were extracted from clinical notes by two means: manually and using an automated clinical natural language processing (CNLP) tool. Finally, 14 previously unsolved cases were reanalyzed. GEM ranked over 90% of the causal genes among the top or second candidate and prioritized for review a median of 3 candidate genes per case, using either manually curated or CNLP-derived phenotype descriptions. Ranking of trios and duos was unchanged when analyzed as singletons. In 17 of 20 cases with diagnostic SVs, GEM identified the causal SVs as the top candidate and in 19/20 within the top five, irrespective of whether SV calls were provided or inferred ab initio by GEM using its own internal SV detection algorithm. GEM showed similar performance in absence of parental genotypes. Analysis of 14 previously unsolved cases resulted in a novel finding for one case, candidates ultimately not advanced upon manual review for 3 cases, and no new findings for 10 cases. GEM enabled diagnostic interpretation inclusive of all variant types through automated nomination of a very short list of candidate genes and disorders for final review and reporting. In combination with deep phenotyping by CNLP, GEM enables substantial automation of genetic disease diagnosis, potentially decreasing cost and expediting case review.

De La Vega FM ，Chowdhury S ，Moore B ，Frise E ，McCarthy J ，Hernandez EJ ，Wong T ，James K ，Guidugli L ，Agrawal PB ，Genetti CA ，Brownstein CA ，Beggs AH ，Löscher BS ，Franke A ，Boone B ，Levy SE ，Õunap K ，Pajusalu S ，Huentelman M ，Ramsey K ，Naymik M ，Narayanan V ，Veeraraghavan N ，Billings P ，Reese MG ，Yandell M ，Kingsmore SF ... -  《Genome Medicine》

Evaluation of an automated genome interpretation model for rare disease routinely used in a clinical genetic laboratory.

The analysis of exome and genome sequencing data for the diagnosis of rare diseases is challenging and time-consuming. In this study, we evaluated an artificial intelligence model, based on machine learning for automating variant prioritization for diagnosing rare genetic diseases in the Baylor Genetics clinical laboratory. The automated analysis model was developed using a supervised learning approach based on thousands of manually curated variants. The model was evaluated on 2 cohorts. The model accuracy was determined using a retrospective cohort comprising 180 randomly selected exome cases (57 singletons, 123 trios); all of which were previously diagnosed and solved through manual interpretation. Diagnostic yield with the modified workflow was estimated using a prospective "production" cohort of 334 consecutive clinical cases. The model accurately pinpointed all manually reported variants as candidates. The reported variants were ranked in top 10 candidate variants in 98.4% (121/123) of trio cases, in 93.0% (53/57) of single proband cases, and 96.7% (174/180) of all cases. The accuracy of the model was reduced in some cases because of incomplete variant calling (eg, copy number variants) or incomplete phenotypic description. The automated model for case analysis assists clinical genetic laboratories in prioritizing candidate variants effectively. The use of such technology may facilitate the interpretation of genomic data for a large number of patients in the era of precision medicine.

Meng L ，Attali R ，Talmy T ，Regev Y ，Mizrahi N ，Smirin-Yosef P ，Vossaert L ，Taborda C ，Santana M ，Machol I ，Xiao R ，Dai H ，Eng C ，Xia F ，Tzur S ... -  《-》

Diagnosis of a Single-Nucleotide Variant in Whole-Exome Sequencing Data for Patients With Inherited Diseases: Machine Learning Study Using Artificial Intelligence Variant Prioritization.

In recent years, thanks to the rapid development of next-generation sequencing (NGS) technology, an entire human genome can be sequenced in a short period. As a result, NGS technology is now being widely introduced into clinical diagnosis practice, especially for diagnosis of hereditary disorders. Although the exome data of single-nucleotide variant (SNV) can be generated using these approaches, processing the DNA sequence data of a patient requires multiple tools and complex bioinformatics pipelines. This study aims to assist physicians to automatically interpret the genetic variation information generated by NGS in a short period. To determine the true causal variants of a patient with genetic disease, currently, physicians often need to view numerous features on every variant manually and search for literature in different databases to understand the effect of genetic variation. We constructed a machine learning model for predicting disease-causing variants in exome data. We collected sequencing data from whole-exome sequencing (WES) and gene panel as training set, and then integrated variant annotations from multiple genetic databases for model training. The model built ranked SNVs and output the most possible disease-causing candidates. For model testing, we collected WES data from 108 patients with rare genetic disorders in National Taiwan University Hospital. We applied sequencing data and phenotypic information automatically extracted by a keyword extraction tool from patient's electronic medical records into our machine learning model. We succeeded in locating 92.5% (124/134) of the causative variant in the top 10 ranking list among an average of 741 candidate variants per person after filtering. AI Variant Prioritizer was able to assign the target gene to the top rank for around 61.1% (66/108) of the patients, followed by Variant Prioritizer, which assigned it for 44.4% (48/108) of the patients. The cumulative rank result revealed that our AI Variant Prioritizer has the highest accuracy at ranks 1, 5, 10, and 20. It also shows that AI Variant Prioritizer presents better performance than other tools. After adopting the Human Phenotype Ontology (HPO) terms by looking up the databases, the top 10 ranking list can be increased to 93.5% (101/108). We successfully applied sequencing data from WES and free-text phenotypic information of patient's disease automatically extracted by the keyword extraction tool for model training and testing. By interpreting our model, we identified which features of variants are important. Besides, we achieved a satisfactory result on finding the target variant in our testing data set. After adopting the HPO terms by looking up the databases, the top 10 ranking list can be increased to 93.5% (101/108). The performance of the model is similar to that of manual analysis, and it has been used to help National Taiwan University Hospital with a genetic diagnosis.

Huang YS ，Hsu C ，Chune YC ，Liao IC ，Wang H ，Lin YL ，Hwu WL ，Lee NC ，Lai F ... -  《-》

Critical assessment of variant prioritization methods for rare disease diagnosis within the rare genomes project.

A major obstacle faced by families with rare diseases is obtaining a genetic diagnosis. The average "diagnostic odyssey" lasts over five years and causal variants are identified in under 50%, even when capturing variants genome-wide. To aid in the interpretation and prioritization of the vast number of variants detected, computational methods are proliferating. Knowing which tools are most effective remains unclear. To evaluate the performance of computational methods, and to encourage innovation in method development, we designed a Critical Assessment of Genome Interpretation (CAGI) community challenge to place variant prioritization models head-to-head in a real-life clinical diagnostic setting. We utilized genome sequencing (GS) data from families sequenced in the Rare Genomes Project (RGP), a direct-to-participant research study on the utility of GS for rare disease diagnosis and gene discovery. Challenge predictors were provided with a dataset of variant calls and phenotype terms from 175 RGP individuals (65 families), including 35 solved training set families with causal variants specified, and 30 unlabeled test set families (14 solved, 16 unsolved). We tasked teams to identify causal variants in as many families as possible. Predictors submitted variant predictions with estimated probability of causal relationship (EPCR) values. Model performance was determined by two metrics, a weighted score based on the rank position of causal variants, and the maximum F-measure, based on precision and recall of causal variants across all EPCR values. Sixteen teams submitted predictions from 52 models, some with manual review incorporated. Top performers recalled causal variants in up to 13 of 14 solved families within the top 5 ranked variants. Newly discovered diagnostic variants were returned to two previously unsolved families following confirmatory RNA sequencing, and two novel disease gene candidates were entered into Matchmaker Exchange. In one example, RNA sequencing demonstrated aberrant splicing due to a deep intronic indel in ASNS, identified in trans with a frameshift variant in an unsolved proband with phenotypes consistent with asparagine synthetase deficiency. Model methodology and performance was highly variable. Models weighing call quality, allele frequency, predicted deleteriousness, segregation, and phenotype were effective in identifying causal variants, and models open to phenotype expansion and non-coding variants were able to capture more difficult diagnoses and discover new diagnoses. Overall, computational models can significantly aid variant prioritization. For use in diagnostics, detailed review and conservative assessment of prioritized variants against established criteria is needed.

Stenton SL ，O'Leary MC ，Lemire G ，VanNoy GE ，DiTroia S ，Ganesh VS ，Groopman E ，O'Heir E ，Mangilog B ，Osei-Owusu I ，Pais LS ，Serrano J ，Singer-Berk M ，Weisburd B ，Wilson MW ，Austin-Tse C ，Abdelhakim M ，Althagafi A ，Babbi G ，Bellazzi R ，Bovo S ，Carta MG ，Casadio R ，Coenen PJ ，De Paoli F ，Floris M ，Gajapathy M ，Hoehndorf R ，Jacobsen JOB ，Joseph T ，Kamandula A ，Katsonis P ，Kint C ，Lichtarge O ，Limongelli I ，Lu Y ，Magni P ，Mamidi TKK ，Martelli PL ，Mulargia M ，Nicora G ，Nykamp K ，Pejaver V ，Peng Y ，Pham THC ，Podda MS ，Rao A ，Rizzo E ，Saipradeep VG ，Savojardo C ，Schols P ，Shen Y ，Sivadasan N ，Smedley D ，Soru D ，Srinivasan R ，Sun Y ，Sunderam U ，Tan W ，Tiwari N ，Wang X ，Wang Y ，Williams A ，Worthey EA ，Yin R ，You Y ，Zeiberg D ，Zucca S ，Bakolitsa C ，Brenner SE ，Fullerton SM ，Radivojac P ，Rehm HL ，O'Donnell-Luria A ... -  《-》

Genome analysis and knowledge-driven variant interpretation with TGex.

Dahary D ，Golan Y ，Mazor Y ，Zelig O ，Barshir R ，Twik M ，Iny Stein T ，Rosner G ，Kariv R ，Chen F ，Zhang Q ，Shen Y ，Safran M ，Lancet D ，Fishilevich S ... -  《BMC Medical Genomics》