119. Concordance of integrated analysis approaches to measure HRD genomic instability

Aly Abdelkareem

Soheil Shams

Dr. Soheil Shams is the Chief Informatics Officer at Bionano Genomics. Prior to this role, he was the founder and president of BioDiscovery, Inc. a privately held bioinformatics company based in El Segundo, California for more than 24 years. He received his doctorate degree from University of Southern California in the field of Computer Engineering, with emphasis on neural networks and parallel processing architectures. He is a leading innovator in genomic informatics, having invented many early approaches to array analysis resulting in numerous U.S. patents. Under the direction of Dr. Shams, BioDiscovery played a pioneering role in development of software tools for microarray and NGS-based research, as well as clinical research applications in cytogenetic and molecular genetic diagnostics. Prior to founding BioDiscovery, he was a Sr. Member of Staff at Hughes Research Laboratories (HRL) in Malibu and taught undergraduate and graduate classes in artificial intelligence, machine perception, and data mining at UCLA. His research interests include the genomics of rare disease and cancer, pattern recognition technologies, and parallel processing computer architectures. He has worked with many leading scientists in genomic research and has authored or co-authored over 80 technical publications and book chapters. Dr. Shams has been recipient of multiple prestigious awards, including the HRL Best Published Paper of the Year award.


Soheil Shamsa, Alina Keomaneea, Raja Kashavana, Megan Roytmana, Daniel Saula, Christopher Lumb

aBioDiscovery, LLC, El Segundo, CA, USA; bDiagnostic Laboratory Service, Inc, Aiea, Hawaii, USA

Homologous Recombination Deficiency (HRD) has been shown to be an effective pharmacogenetic biomarker for determining the efficacy of PARP inhibitor therapy across various tumor types. There are two common approaches to measuring the functional genomic instability associated with HR deficiency by either a) calculating the total genomic Loss of heterozygosity (gLOH), or b) measuring three defined genomic scars: Loss of heterozygosity (LOH), Telomeric Allelic Imbalance (TAI), & Large-Scale State Transitions (LST).

Previously, we curated results from The Cancer Genome Atlas (TCGA) to produce high fidelity copy number information by correcting for over-segmentation and incorrect ploidy for many samples provided publicly through the TCGA website as level 3 data. Our manually curated data is referred to as TCGA Premier. We analyzed 529 ovarian cancer samples from TCGA Premier with the automated scarring approach implemented in NxClinical v6.2 and categorized each sample based on its number of detected scars. A subset of the samples had previously reported HRD scores (Takaya et al., 2020), which were used to establish validity of the automated scarring results as well as to generate overall concordance. In addition, a comparative assessment of the genomic scarring approach to the total genomic LOH approach was conducted on the full dataset. Insights on the concordance between the two genomic stability analysis modalities will be provided, as well as case studies highlighting the importance of having a complete picture of the tumor profile.