Poster M9 - Automated Oncoprint Generation from Public Genomic Data

Introduction: Oncoprints are sophisticated visual representations showcasing the complexity of genomic, along with occasional transcriptomic or proteomic, alterations found within cancer patient cohorts. These graphical summaries condense the spectrum of genetic changes across individuals, facilitating the interpretation of expansive sequencing data sets by scientists and clinicians. Beyond their summarizing capability, oncoprints possess the versatility to underscore the mutual exclusivity of genetic alterations among diverse patient samples and across a range of published studies, providing an expansive view of the cancer genomic landscape. The strategic use of oncoprints further facilitates the field of precision oncology by unveiling novel genetic targets for tailored drug therapies. They possess the capability to summarize the treatment efficacy and the emergence of resistance mechanisms through comparative analyses of pre- and post-therapeutic sequencing data. Most importantly, the frequent application of oncoprints in the pursuit of biomarker discovery constitutes the advancement of precise diagnostic and therapeutic stratagems for a variety of cancer types.

The process of Oncoprint creation can pose a complex, time-consuming task for scientists and clinicians without computational skill or experience. This abstract outlines and presents the progress toward building an automated Oncoprint generation tool that can be utilized to visualize complex genomic data from public datasets with specified cancer types being the only.

Materials and

Methods: Data Acquisition and Pre-processing:
Clinical sample data from the openpedcan v12 dataset was imported in R with the fread function from data.table, focusing on samples with pediatric brain-related cancers. Specimen identifiers were extracted for subsequent data amalgamation.

Data Integration and Mutation Matrix Composition:
A comprehensive cancer gene list was compiled from multiple TSV files. Our central mutation dataset was filtered to contain only genes from this list. A matrix structure was formed, arranged with rows as genes and columns as patient IDs, to represent the mutation data. An iterative approach filled in the matrix based on variant classifications from the mutation data.

Oncoprint Construction:
We established a color scheme to visually differentiate mutation types using the ComplexHeatmap package. Corresponding graphical functions were defined for each mutation type to represent alterations accurately. Filtering criteria were applied to maintain rows and columns with significant mutation frequency.

Heatmap Visualization:
We generated the oncoprint by integrating the mutation matrix, color scheme, and graphical functions. The sample oncoprint centered on pediatric brain-related cancers was generated through this program, highlighting the distribution of alterations.

Results, Conclusions, and Discussions: The R pipeline successfully automated the oncoprint creation process. When utilizing specific cancer types, the program generated oncoprints showcasing genes with mutations present for patients with the specified cancers. Broadening the analysis to encompass all types of brain cancers revealed patterns of mutual exclusivity and similarity between genes. The oncoprints also revealed varying mutation burdens across patients. Our automated approach for classifying mutations based on actionability refined the final visual representation, ensuring only clinically significant alterations were highlighted.

The implementation of our R-based oncoprint pipeline offers a novel and customized approach to visualizing genomic alterations in cancer patient cohorts with user inputs being simply cancer types. It addresses previous limitations by allowing the Euclidean sorting and comparison of actionable alterations and simplifies the integration of various patient groups. Our findings reinforce the utility of oncoprints in identifying patterns of mutual exclusivity, which is critical for precision oncology. Furthermore, the automated segmentation of actionable mutations improves the relevance of the output for clinical and research purposes. Future steps include the expansion of this Oncoprinter for an extensive pan-cancer analysis of pediatric patients and the incorporation of transcriptomic and proteomic data. The creation of an internal dataset for actionable aberrations aims to further enhance the pipeline's utility, supplemented by the use of an R AI package to resolve formatting inconsistencies.

Acknowledgements (Optional):