Estimating the order of mutations during tumorigenesis from tumor genome sequencing data

Abstract

Motivation: Tumors are thought to develop and evolve through a sequence of genetic and epigenetic somatic alterations to progenitor cells. Early stages of human tumorigenesis are hidden from view. Here, we develop a method for inferring some aspects of the order of mutational events during tumorigenesis based on genome sequencing data for a set of tumors. This method does not assume that the sequence of driver alterations is the same for each tumor, but enables the degree of similarity or difference in the sequence to be evaluated.

Results: To evaluate the new method, we applied it to colon cancer tumor sequencing data and the results are consistent with the multi-step tumorigenesis model previously developed based on comparing stages of cancer. We then applied the new method to DNA sequencing data for a set of lung cancers. The model may be a useful tool for better understanding the process of tumorigenesis.

Availability: The software is available at: http://linus.nci.nih.gov/Data/YounA/OrderMutation.zip.

Fig. 1.

Distribution of mutations in driver genes and estimates of their order (a) Distribution of mutations in four driver genes for ten patients of the same tumor type. For each patient, the colored letter ‘M’ represents a mutation for the corresponding gene. (b) Estimates for P_k,i, the probability that the k-th mutational event involving the four driver genes occurs in gene i. The length of the sub-bar corresponding to gene i at the k-th mutational event is the estimates for P_k,i

Fig. 2.

Most frequently mutated genes at each mutational step for non-small-cell lung tumors (Ding et al., 2008). The length of the sub-bar corresponding to gene i at the k-th mutational step is the MLE of P_k,i

Fig. 3.

Distribution of mutational step for frequently mutated genes in non-small-cell lung tumors (Ding et al., 2008) The height of the bar at the k-th mutational step for gene i is the MLE of P_k,i

Fig. 4.

Most frequently mutated genes at each mutational step for colorectal tumors (Wood et al., 2007). The length of the sub-bar corresponding to gene i at the k-th mutational step is the MLE of P_k,i

Fig. 5.

Distribution of mutational step for frequently mutated genes in colorectal tumors (Wood et al., 2007). The height of the bar at the k-th mutational step for gene i is the MLE of P_k,i