Mutually exclusive mutation profiles define functionally related genes in muscle invasive bladder cancer

Abstract

Muscle Invasive bladder cancer is known to have an abundance of mutations, particularly in DNA damage response and chromatin modification genes. The role of these mutations in the development and progression of the disease is not well understood. However, a mutually exclusive mutation pattern between gene pairs could suggest gene mutations of significance. For example, a mutually exclusive mutation pattern could suggest an epistatic relationship where the outcome of a mutation in one gene would have the same outcome as a mutation in a different gene. The significance of a mutually exclusive relationship was determined by establishing a normal distribution of the conditional probabilities for having a mutation in one gene and not the other as well as the reverse relationship for each gene pairing. Then these distributions were used to determine the sigma-magnitude of standard deviation by which the observed value differed from the expected, a value that can also be interpreted as the 'p-value'. This approach led to the identification of mutually exclusive mutation patterns in KDM6A and KMT2D as well as KDM6A and RB1 that suggested the observed mutation pattern did not happen by chance. Upon further investigation of these genes and their interactions, a potential similar outcome was identified that supports the concept of epistasis. Knowledge of these mutational interactions provides a better understanding of the mechanisms underlying muscle invasive bladder cancer development, and may direct therapeutic development exploiting genotoxic chemotherapy and synthetic lethality in these pathways.

Fig 1. Pipeline for dataset generation.

Pathways show the data selection and trimming process through each of the Lest Conservative, Mid Conservative and Most Conservative approaches. The number of cases, genes and mutations (mut) in the data at each step of the refinement process are presented in square boxes.

Fig 2. Coefficients of variation (CV) for the cases of MIBC represented in the different mutation annotation file (MAF).

Fig 3

Visualization of the overlap in mutation calling between the four pipelines (Muse, Mutect2, SomaticSniper, Varscan2) mutation calling pipelines in four cases with high CV (a-d), and two cases with low CV. Sample details are: (A) TCGA-XF-A9SL-01A-11D-A391-08, CV = 1.3, 35 mean / 100 total mutation calls; (B) TCGA-K4-A83P-01A-11D-A34U-08, CV = 0.98, 125 mean /317 total mutation calls; (C) TCGA-XF-A9SI-01A-11D-A391-08, CV = 0.71, 215 mean / 450 total mutation calls; (D) TCGA-DK-A1AF-01A-11D-A13W-08, CV = 1.00, 45 mean / 120 total mutation calls; (E) TCGA-LT-A8JT-01A-11D-A364-08, CV = 0.026, 90 mean / 100 total mutation calls; (F) TCGA-DK-A6AW-01A-11D-A30E-08, CV = 0.020, 5000 mean / 5300 total mutation calls.

Fig 4. Representative differences in statistical significance of conditional probability for different gene pairs.

The red bar is the mean of the RGCP distribution, and the blue bar is the value observed in the CP matrix. (A) A relationship where there is no statistically significant difference between RGCP distribution and the observed value, between the KDM6A mutation and no mutation in EP300. (B) A statistically significant relationship, where RGCP distribution for having KDM6A mutation and no mutation in KMT2D is significantly different from the observed value. (C and D) A statistically significant, bidirectional relationship, where the presence of an RB1 mutation and absence of a KDM6A mutation, and the (opposite) absence of an RB1 mutation and presence of a KDM6A mutation are both observed at frequencies significantly different from the predicted RGCP distribution.

Fig 5. Summary of mutually exclusive relationships identified in the study.

Lines between the identified genes indicate which data subset(s) the relationships were identified in, as well as the statistical strength of the observations. DDR genes are indicated by square frames, CM genes by oval frames, and genes overlapping both processes in rectangular frames with round corners. Interactions are shown for: Least High and Moderate Impact Mutations (red lines), Least High Mutations only (blue lines), Mid High and Moderate Impact Mutations (black lines), and Most High and Moderate Mutations (green lines). Statistical significance indifferences between RGCP and mean observed values is presented as: Sigma > 2.7 (double line), Sigma >2 (single solid line), Sigma >1.5 (dashed line).