Evaluation of transcriptome forward computational strategies to improve molecular diagnosis in rare pediatric neuromuscular disease

Topic:

Other

Poster Number: 173

Author(s):

Sarah Silverstein, NNDCS/NINDS/NIH, Sandra Donkervoort, Genetic Counselor, Neuromuscular and Neurogenetic Disorders of Childhood Section, NINDS, NIH, Justin Moy, Bioinformatics Program, Boston University, Brian Uapinyoying, PhD, NNDCS/NINDS/NIH, Kaushik Ganapathy, Scripps Research Institute of La Jolla, Svetlana Gorokhova, MD, PhD, Department of Medical Genetics, Timone Childrens Hospital Marseille, Thomas DeLong, Neuromuscular and Neurogenetic Disorders of Childhood Section, NINDS, NIH, Vijay Ganesh, MD, PhD, Center for Mendelian Genomics, Broad Institute of MIT, Ben Weisburd, PhD, Broad Institute of MIT, Rotem Orbach, MD, NNDCS/NINDS/NIH, A. Reghan Foley, MD, Neuromuscular and Neurogenetic Disorders of Childhood Section, NINDS, NIH, Pejman Mohammadi, PhD, Scripps Research Institute of La Jolla, David Adams, MD, PhD, NHGRI/NIH, Carsten Bönnemann, MD, Neuromuscular and Neurogenetic Disorders of Childhood Section, NINDS, NIH

Pediatric neuromuscular diseases are a genetically and clinically heterogenous group of disorders of which 30-60% remain molecularly undiagnosed after evaluation with exome or genome sequencing. RNA studies are an emerging tool to augment genome analysis and improve diagnosis by detecting single allele transcription, interpreting potential aberrant splicing and identifying outlier transcription levels. Adding transcriptome analysis to the diagnostic pipeline may increase the diagnostic rate by 16-32% amongst specific disease groups. Recently, open-source computational tools to systematically survey transcriptomic sequencing for aberrant events have become available. In this study we assess the sensitivity of DROP, MINTIE, LeafcutterMD, rMATS-turbo and HAPASE/ANEVA-H to detect known outliers in pediatric neuromuscular cohorts in order to identify the best tools for diagnostic use. Each tool selected utilizes a unique statistical strategy and experimental design.

Each tool is to be run on three transcriptome cohorts from the NINDS Neuromuscular and Neurogenetic Disorders of Childhood Section: muscle biopsy (n=71), cycloheximide treated skin fibroblasts (n=79) and untreated skin fibroblasts (n=37). All samples were prepared by poly-A selection or ribo-depletion library preparation methods, sequenced at 50-300 million paired reads and aligned using the GTEXv10 pipeline. Principle component analysis helped exclude low quality samples and construct homogenous cohorts. Final samples were manually curated for true positive diagnoses that should be detectable by each tool.

The DROP pipeline correctly identified 4/28 (14%) known diagnoses and yielded one new candidate diagnosis of TRIP4-related disease that is currently being validated. Preliminarily, the DROP pipeline performs better on fibroblast cohorts than muscle cohorts. The remaining tools are currently undergoing assessment for sensitivity of detection and new candidate diagnoses. Based on the forthcoming results, recommendations will be made for the use of these tools to improve diagnosis in rare pediatric neuromuscular disease.