Clinical Assessment of Next-Generation Sequencing Probe Reproducibility in Short-Read Sequencing (ClinRay) Using Digital Twins

Abstract

Short-read next-generation sequencing is widely used for clinical diagnosis but faces limitations in accurately detecting variants in complex genomic regions, such as segmental duplications, guanine-cytosine rich areas, and repeat sequences. These challenging regions comprise only 3% to 5% of the exome, yet their stochastic library preparation and bioinformatics analysis impacts variant detection reproducibility. Evaluating reproducibility is time-consuming, and variants in these regions require validation through sequencing replicates and using orthogonal methods like long-range PCR or Sanger sequencing, thereby increasing costs and turnaround times for clinical laboratories. To address these challenges, ClinRay was developed; it is a novel, generalizable bioinformatics method that uses the concept of digital twins to synthetically enhance the data distribution for variants in regions with suspected poor reproducibility. ClinRay predicts the reproducibility of detected variants by short-read next-generation sequencing probes in these difficult-to-sequence genomic regions. The model was developed using alignment data from the binary format of the sequence alignment/map files of eight replicates of the Genome in a Bottle HG002 Coriell cell and publicly available genomic context annotation resources. The model achieved an area under the receiver-operating characteristic curve of 0.89 (95% CI, 0.88–0.90) on the test data and 0.85 (95% CI, 0.84–0.86) on an independent validation data set.