(P 26) SARCDBASE: A TOOL FOR DETECTION OF GENETIC ALTERATIONS IN SARCOMA

Objective: Sarcomas are a diverse group of malignant tumors arising in the bone and soft tissues. For an increasing number of subtypes, there are already pathognomonic genetic alterations described. For some sarcomas, such alterations remain to be identified. The latter group primarily contains sarcomas with a massive amount of genomic copy number and structural alterations. The advent of high-throughput sequencing techniques allowed for the detection of basically all mutations in cancer biopsies, but for high-grade cancers the number of mutations is staggering. This makes it difficult to extract biologically significant ones from background noise. We have developed SarcDBase, a database-building method that automatically matches genetic variants detected in high-resolution genomic and transcriptomic data from tumor biopsies with information on established biomarkers, eliminating the need for prior knowledge and manual screening of data. The output provides detailed information on both established biomarkers and novel variants with probable biological significance. Additionally, it provides diagnostic predictions.

Methods: The tool is divided into three main sections. 1. Data Analysis: DNA and RNA extracted from fresh frozen tumor biopsies are sequenced to detect (i) structural variants and gene fusions, (ii) single nucleotide variants and insertions/deletions and (iii) genome-wide DNA copy numbers changes. The bioinformatics analysis also comprises NAFuse (https://github.com/ValeriaDifilippo/NAFuse), an integrated methodology that combines and compares the output of a gene fusion detector with that of the BAM-file generated by whole genome sequencing. 2. SarcDBase Execution: a series of codes that use external sources, such as the WHO Classification of Tumours (Soft Tissue and Bone Tumours), gene fusions curated from a targeted literature search, and the Mitelman Database of Chromosome Aberrations and Gene Fusions in Cancer, to detect previously reported genetic alterations and new ones of likely biological significance, e.g., new combinations of partner genes in gene fusions. 3. SarcDBase Visualization: an app in dash that helps the user to visualize and interpret the data with a graphical interphase and to provide a dynamic visualization. The performance of SarcDBase was tested on a training cohort (n = 46) where the analyses were blinded. Based on the findings obtained, modifications were made, and a validation cohort (n = 114) was then used to reassess the tool’s performances.

Results: Several mutations of likely biological and diagnostic significance were identified. Among these, we found mutations in some genes reported in the WHO book that confirmed the diagnosis, such as the NAB2::STAT6 gene fusion, typical of solitary fibrous tumor, the JAZF1::PHF1 in endometrial stromal sarcoma, and the H3F3A p.G35W in giant cell tumor of bone. SarcDBase was also able to identify new genetic alterations of likely biological significance, such as new combinations of partner genes in gene fusions involving, for example, the TP53 gene.

Conclusion: SarcDBase is a useful tool to integrate information on genomic copy numbers, transcriptome, single nucleotide, and structural variants, and to extract previously reported alterations and new ones of likely biological importance. It can be used on any samples with deep sequencing data, and it also provides graphical visualization to help the user evaluate the findings. The user-friendly interface significantly reduces the entry barrier for users in terms of bioinformatics competence. It automatically links genetic data from public sources to specific alterations found in tumor samples, eliminating the need for user expertise or manual data sifting. However, patient information, tumor location, histomorphology, and other characteristics are always required to contextualize the genetic findings and to enhance SarcDBase’s performance. In rare tumors such as sarcomas, more knowledge must be gained in order to properly train developed tools to assess what sarcoma subtype is at hand.