Abstract Somatic genetic aberrations are a fundamental driver of cancer. Their characterization in cancer genomes is increasingly being used to guide therapeutic decisions. The ongoing evolution of subclones in a growing tumor leads to a heterogeneous and complex mixture of mutations and other genetic aberrations with a wide range of different allelic fractions. Accurately deciphering a tumor's clonal composition and structure frequently requires very high genome sequencing coverage, in the thousand-fold range if not higher, to reliably detect and quantify variant mutant alleles. However, current short read sequencing technologies are unable to distinguish individual DNA molecules from duplicates without the use of error-prone and complicated molecular indexing strategies. Thus, the vast majority of cancer genome sequencing data is confounded by limitations on linking cancer genome DNA sequences to their clonal origin. In this study, we developed a new and robust strategy to identify individual DNA molecule sequences in cancer genomes, improve detection of mutations at minor allelic fractions, and thus reliably delineate clonal populations in tumors. We introduce a unique and highly error-resistant DNA molecular tag that can identify every individual DNA molecule in the sample. This molecular tag is based on the combinatorial assembly of rationally-designed sequences and utilizes engineering principles that minimize data transmission errors. It readily scales from billions to trillions of molecular species without prior sequence knowledge and is easily incorporated using a rapid transposase-based sequencing library protocol. In stark contrast, existing molecular indexing methods require random nucleotide DNA barcodes to ostensibly quantify molecular species. Unlike our new combinatorial molecular tagging system, the older random barcode method is highly sensitive to sequencing errors that substantially reduces performance in clonal delineation. As a proof-of-concept, we assessed the quantification of variants and gene isoforms respectively from known admixtures of well-characterized diploid genomes and mRNA samples. We determined that high depth sequencing leads to hidden but extensive variability in variant quantification, but can be corrected with our approach. Subsequently, we applied our combinatorial DNA molecular tagging technology to matched normal-primary tumor samples. By performing ultra-high depth sequencing for approximately 100 cancer-related genes, we can identify mutations at single molecule resolution and can determine clonal composition. We also observed substantial variability in somatic variant abundances that was previously undetectable. Overall, we demonstrated an approach to study the clonal structure in cancer genomes through the identification and characterization of individual molecules. Citation Format: Billy Lau, Hanlee Ji. Clonal structure analysis of cancer genomes at single molecule resolution. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4889. doi:10.1158/1538-7445.AM2015-4889