Characterization of copy number variations (CNV) patterns and pseudotime trajectories in high-grade serous ovarian cancer (HGSOC).

5595 Background: HGSOC is characterized by genomic instability, resulting in frequent CNVs. This study aimed to characterize CNV patterns and derive pseudotime trajectories in HGSOC. Methods: Patients with a diagnosis of HGSOC enrolled in the CGP program (NCT06020625) at Fondazione Policlinico Universitario A. Gemelli between March 2022 and December 2023 were analyzed using the TruSight Oncology 500 (TSO500) platform, covering CNVs across 523 genes. CNVs were classified as high- or low-level based on log2 segmented CN profiles. An absolute log2 value of 0.3 was used to define the fraction of altered genes. High-variability genes were identified using a variance threshold. The Leiden algorithm was applied to identify clusters based on modularity, and connectivity structures were mapped using the partition-based graph abstraction (PAGA) method. Pseudotime trajectories were modeled using diffusion-like random walks, with the root cluster chosen based on the fraction of altered genes. Non-negative matrix factorization (NMF) was used to identify gene-level CNV patterns. The number of altered genes per region was normalized to the total number of genes per chromosome arm. Chi-squared test was used to compare categorical variables. Results: A total of 597 HGSOC patients, 91% of whom were FIGO stage III-IV, were included. TP53 mutations were identified in 96% of patients. MYC (13%) and CCNE1 (9%) were the most frequently high-level amplified genes. 104 highly variable genes were included in the trajectory analysis. The Leiden algorithm identified 7 patient clusters, with a significant trend observed across clusters in the fraction of altered genes (p<0.001, Kendall’s test for trend). Clusters 2 and 3 showed the lowest fraction of altered genes and were the most connected in the PAGA analysis. Using cluster 3 as the root, trajectory analysis revealed divergent branching patterns, with a significant Spearman correlation between pseudotime and the fraction of altered genes (ρ=0.20, p<0.001). Clusters also showed different percentages of homologous recombination (HR)-proficient patients (p<0.001) and mutations in BRCA1 (p=0.02) and BRCA2 (p=0.006). NMF identified 4 gene clusters based on CNV patterns, with distinct amplification and deletion profiles. Each cluster showed a different ratio of low/high-level CNVs. In the cluster characterized by high-level amplifications, CCNE1 and MYC low/high-level ratios were 1.1 and 2.0, respectively. While specific chromosomal regions were enriched in the cluster characterized by low-level CNVs (p<0.001), clusters with high-level CNVs lacked specific regional associations. Conclusions: This study suggests that distinct patterns of CNVs in HGSOC are linked with genomic and clinical features. This application of CGP for CNV profiling could offer a cost-effective strategy to stratify HGSOC patients and guide personalized treatments. Clinical trial information: NCT06020625 .