Abstract PHB03: A hierarchical model of DNA repair inferred from omics-scale genetic interaction data reveals the dynamics of DNA damage induction

Deficiency in one or more subsystems of the DNA damage response and repair system hierarchy is a hallmark of cancer and makes DNA repair an attractive target for physical or chemical therapies. An accurate, complete, yet parsimonious model of the systems and pathways comprising the hierarchy of DNA repair mechanisms can be expected to further our understanding of cancer genesis and progression and to facilitate the development of improved cancer therapies. Previous attempts in building a hierarchical model of DNA repair include methods based on literature curation (Gene Ontology [1]) and expert consensus (2). However, as these models rely on manual curation, rapid integration of novel omics-scale experimental data sets is not practicable. Moreover, Gene Ontology is by its definition only concerned with the modeling of pathways in the healthy cell and does not aim to reflect cancer-specific aspects. To bridge this gap, we here describe the generation of an entirely data-driven model of DNA repair, inferred from a large meta-compendium describing the physical interactions between proteins based on an integration of over 110 experimental data sets from different omics levels (genome, transcriptome, proteome). Augmenting a technique that we developed earlier, Active Interaction Mapping (AIM) (3), we define a hierarchical model of DNA repair consisting of 293 genes organized in 40 systems. Briefly, we integrate the interaction evidence using a random forest regressor to generate a weighted, integrated network; we then transform the network into a consensus matrix and feed this into an algorithm based on previous work (4) to identify the hierarchical structure embedded in the network. The hierarchical model takes the form of a directed acyclic graph (DAG), which allows us to model pleiotropic aspects of molecular systems, as some proteins and systems play functional roles in different ancestor systems. Using a network biology approach, we suggest a list of 293 genes playing central roles in DNA repair and find evidence that genome replication and RNA splicing processes are more intimately connected to DNA repair than previously appreciated. We find novel roles for about one third of the genes in our model, identify novel subsystems of already known systems in DNA repair, and can describe novel interactions between already known DNA repair systems. Finally, we perform perturbations of the DNA repair hierarchy in different cell lines, measure the impact on the protein-protein and genetic interaction networks, and use the identified interactions to search for impacted connections within and between systems of the DNA repair model. References: 1. Ashburner M et al., Nature Genetics 2000;25:25. 2. Pearl LH et al., Nature Reviews Cancer 2015;15:166. 3. Kramer MH et al., Molecular Cell 2017;65:761. 4. Kramer M, Dutkowski J, Yu M, Bafna V, Ideker T. Bioinformatics 2014;30:i34. This abstract is also being presented as Poster B43. Citation Format: Anton Kratz, Fan Zheng, Min Kyu, Nevan Krogan, Trey Ideker. A hierarchical model of DNA repair inferred from omics-scale genetic interaction data reveals the dynamics of DNA damage induction [abstract]. In: Proceedings of the AACR Special Conference on the Evolving Landscape of Cancer Modeling; 2020 Mar 2-5; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2020;80(11 Suppl):Abstract nr PHB03.