While there have been notable advancements in structural variant analysis in tumor genomes, there is still a lack of comprehensive analysis of tumors integrating multiple technologies to fully characterize events not previously discovered using current tools. In this study, we have applied a non-sequencing-based genome mapping technology, Bionano, together with 10x Genomics linked-reads and standard whole genome sequencing to three gastric cancer, two brain cancer and one gastrointestinal stromal tumor (GIST) samples. Leveraging the Bionano technology, we captured structural variants including those at low frequency by combining assembly and rare Bionano SV pipelines. We identified 24 sample-specific variants in brain cancer samples, 90 sample-specific variants in GIST and average of 231 sample-specific variants in gastric cancer. Additionally, we present results of our newly developed tools (i) Novel-X, a method for assembly of novel sequence insertions in linked-read sequenced genome which is based on a local assembly of interconnecting barcodes, and (ii) VALOR2, a mapping based algorithm for characterizing complex large-scale SVs from 10x linked-reads on this cohort. We show that our integrated approach is able to identify many more variants than one single approach (an average increase of 400 inversions over any one method, for example), and characterize the composition of structural variants for each sample. Valor2 is able to identify an average of 80 variants across all samples, a majority of which is not identified by longranger. We also identified deletions which affect PSD3 – a candidate tumor suppressor gene. The combination of Bionano and 10x linked-read with data from standard WGS of tumor-normal pairs allows somatic events to be identified within these samples. We present a comprehensive set of variations in gastric and brain cancer ranging from simple events to large and complex events such as reciprocal and inverted-reciprocal translocations in order to identify clinically relevant events.