P735: Machine learning-guided GWAS in the Taiwan biobank reveals novel height-associated genes with possible epistatic interactions

specimens often yields acceptable results, this incurs substantial time and cost.Here, we sought to evaluate how incorporating a custom prenatal reference file in CMA analysis could improve data quality and optimize downstream analysis.Methods: A representative set of 285 direct prenatal CMA cases (142 AF, 132 POC, and 11 CVS) previously performed at UCLA from 2021 to 2023 were retrospectively selected for reference file generation using Affymetrix ChAS software.Inclusion criteria were limited to cases with no reportable variants and passing all established CMA QC metrics, with an effort to balance both sex and ethnicity.A mutually exclusive set of prenatal cases was assembled to assess the performance of this prenatal reference file on downstream analysis, by comparing CMA assay QC metrics and number of calls to those generated using our existing reference based on blood specimens from 200 normal individuals.To assess clinical sensitivity, we reprocessed and reanalyzed 94 cases with previously reported variants.Results: Data analysis using prenatal-specific reference resulted in a dramatic reduction of noise signal and lower the number of artifact calls significantly.The average number of calls/samples was reduced by 21% (24.5 to 19.4), and this improvement was greatest for POCs, showing a 30% reduction (24.7 to 17.2).Even more strikingly, approximately 65% of samples deemed unsuitable for analysis, which is likely due to overwhelming artifact calls, resulted in acceptable or even clean data simply through reprocessing with the prenatal reference file.Among direct prenatal specimens already suitable for analysis, the prenatal reference outperformed blood-based reference in every established CMA QC metric.These improvements came without a loss of sensitivity, as we achieved 100% clinical concordance for prenatal specimens with previously reported variants, both direct and cultured.Conclusion: For prenatal CMA testing on the Affymetrix platform, incorporation of a custom reference file based on a representative set of direct specimens is an effective and cost-efficient way to improve CMA call quality, reduce repeat testing, improve turnaround time, and lower the overall burden of interpretation.