Paralog dispensability shapes homozygous deletion patterns in tumor genomes

Incomplete paralog compensation generates selective dependency on TRA2A in cancer

Paralogs often exhibit functional redundancy, allowing them to effectively compensate for each other's loss. However, this buffering mechanism is frequently disrupted in cancer, exposing unique paralog-specific vulnerabilities. Here, we identify a selective dependency on the splicing factor TRA2A. We find that TRA2A and its paralog TRA2B are synthetic lethal partners that function as widespread and largely redundant activators of both alternative and constitutive splicing. While loss of TRA2A alone is typically neutral due to compensation by TRA2B, we discover that a subset of cancer cell lines are highly TRA2A-dependent. Upon TRA2A depletion, these cell lines exhibit a lack of paralog buffering specifically on shared splicing targets, leading to defects in mitosis and cell death. Notably, TRA2B overexpression rescues both the aberrant splicing and lethality associated with TRA2A loss, indicating that paralog compensation is dosage-sensitive. Together, these findings reveal a complex dosage-dependent relationship between paralogous splicing factors, and highlight how dysfunctional paralog buffering can create a selective dependency in cancer.

BaCoN (Balanced Correlation Network) improves prediction of gene buffering

Buffering between genes, where one gene can compensate for the loss of another gene, is fundamental for robust cellular functions. While experimentally testing all possible gene pairs is infeasible, gene buffering can be predicted genome-wide under the assumption that a gene's buffering capacity depends on its expression level and its absence primes a severe fitness phenotype of the buffered gene. We developed BaCoN (Balanced Correlation Network), a post hoc unsupervised correction method that amplifies specific signals in expression-vs-fitness correlation networks. We quantified 147 million potential buffering relationships by associating CRISPR-Cas9-screening fitness effects with transcriptomic data across 1019 Cancer Dependency Map (DepMap) cell lines. BaCoN outperformed state-of-the-art methods, including multiple linear regression based on our compiled gene buffering prediction metrics. Combining BaCoN with batch correction or Cholesky data whitening further boosts predictive performance. We characterized 808 high-confidence buffering predictions and found that in contrast to buffering gene pairs overall, buffering paralogs were on different chromosomes. BaCoN performance increases with more screens and genes considered, making it a valuable tool for gene buffering predictions from the growing DepMap.

Machine-learning analysis reveals an important role for negative selection in shaping cancer aneuploidy landscapes

BACKGROUND

Aneuploidy, an abnormal number of chromosomes within a cell, is a hallmark of cancer. Patterns of aneuploidy differ across cancers, yet are similar in cancers affecting closely related tissues. The selection pressures underlying aneuploidy patterns are not fully understood, hindering our understanding of cancer development and progression.

RESULTS

Here, we apply interpretable machine learning methods to study tissue-selective aneuploidy patterns. We define 20 types of features corresponding to genomic attributes of chromosome-arms, normal tissues, primary tumors, and cancer cell lines (CCLs), and use them to model gains and losses of chromosome arms in 24 cancer types. To reveal the factors that shape the tissue-specific cancer aneuploidy landscapes, we interpret the machine learning models by estimating the relative contribution of each feature to the models. While confirming known drivers of positive selection, our quantitative analysis highlights the importance of negative selection for shaping aneuploidy landscapes. This is exemplified by tumor suppressor gene density being a better predictor of gain patterns than oncogene density, and vice versa for loss patterns. We also identify the importance of tissue-selective features and demonstrate them experimentally, revealing KLF5 as an important driver for chr13q gain in colon cancer. Further supporting an important role for negative selection in shaping the aneuploidy landscapes, we find compensation by paralogs to be among the top predictors of chromosome arm loss prevalence and demonstrate this relationship for one paralog interaction. Similar factors shape aneuploidy patterns in human CCLs, demonstrating their relevance for aneuploidy research.

CONCLUSIONS

Our quantitative, interpretable machine learning models improve the understanding of the genomic properties that shape cancer aneuploidy landscapes.