Project Summary An extensible analysis platform will be developed to accurately perform the automated genotyping of PCR/capillary electrophoresis (CE) traces for multiple disease-associated short tandem repeater (STR) assays. This study will evaluate the feasibility of developing generalizable and adaptive molecular analysis models, and will ultimately establish a new paradigm for deep learning analytical tools in the molecular diagnostic space. Advanced machine learning strategies will be applied to interpret genotypes of inherited disorders caused by genetically unstable STR DNA sequences. STRs have traditionally been difficult to investigate due to their length (on the order of kilobases) and low sequence complexity, which elude detection by traditional and next- generation sequencing technologies. However, advances in PCR/CE technology have enabled the amplification and fragment sizing of STR DNA fragments, advancing clinical research and diagnostic test development for several neurodegenerative disorders, such as fragile X syndrome and amyotrophic lateral sclerosis. Despite these advances, the analysis of PCR/CE data from assays targeting STRs remains a manual, burdensome, and subjective process. There is a clear need to create a system that can scale with the development of new assays, and the proposed approach utilizes modern breakthroughs in artificial intelligence to fulfill that need. This method will leverage recent advances in representation learning to establish a generalized and adaptive framework for automated PCR/CE annotation that can scale to new assays and improve automatically with the inclusion of new data. The project will benefit from Asuragen?s experience in optimizing repeat-primed chemistries to develop and commercialize multiple high performance assays including the AmplideX PCR/CE FMR1 kit. Importantly, the proposed modeling strategy will borrow-strength across multiple established PCR/CE assays and generalize to future PCR/CE assays for novel STR disease associated biomarkers. This system will be paramount to enabling a continuous learning platform wherein computationally-assisted annotation of PCR/CE assays can be continuously improved and integrated in to clinical research tools and diagnostics.