Web tools for physician-driven diagnostic interpretation of genomic patient data

点击翻译按钮获取中文摘要

SUMMARY/ABSTRACT Genomic sequencing provides definitive disease diagnoses for many patients with suspected genetic disease, ending or preventing lengthy and costly diagnostic odysseys. However, despite extensive efforts of research clinicians and all current computational analysis technologies, the genetic cause of disease remains unresolved for over half of the sequenced patients in genetics clinics today. All too often, diagnosis from whole- exome or genome sequencing data remains elusive even for patients suffering from diseases with well- understood clinical presentation and genetic architecture. Although diagnostic failure can have multiple causes, we hypothesize that two reasons contribute significantly. First, current variant prioritization tools work by reductive filtering on annotations and inheritance patterns to reduce sets of exonic or genomic variants to small, prioritized lists of candidates. This approach works when clear causative variants are present, but offers minimal capacity to remove highly ranked but false positive candidates, and provides little guidance when causative variants have been missed, typically because of unrecognized data quality problems such as low sequence coverage or exon dropouts. When the first round of analysis yields no plausible candidates, current tools don't have the ability to suggest a sensible “next step”, e.g. to deepen or expand the search for causative variants in the data, and the result is analysis dead-end. Second, because of onerous IT expertise and bioinformatic skill requirements, physicians currently rely on bioinformatics experts to analyze genomic data. However, the bioinformatician does not possess the physician's clinical expertise or detailed knowledge of disease presentation, clinical phenotype, and time course of the disease, all of which can be critical in making a diagnosis. This gap between clinical and computational expertise hinders diagnostic success and disease discovery. Here we propose to build a set of web tools that offer novel functionality for deeper, systematic re- examination of the data for disease-causing variants, but are also intuitive and easy to use so clinicians can themselves analyze their patients' genomic datasets. These tools will be based on our already popular IOBIO system available at http://iobio.io, and will offer diagnostic analysts the ability to rapidly examine the quality of their genomic datasets, and visually and in real time search the patient's data for disease causing variants. A unique aspect of this development is that it will be physician-driven from the outset: a large team of clinical Investigators will help design, prioritize, and evaluate software features, and integrate the tools into physician practice and training, ensuring these tools will be usable by clinicians, and they address the most relevant analysis steps for successful clinical diagnosis. Our tools and training materials will be made widely available, drastically lowering the barrier to participation in genomic data analysis for all clinicians who can benefit from genomic data of their patients, helping genomic sequencing to reach its potential as a means to make definitive clinical diagnoses.