Mathematical modeling the order of driver gene mutations in colorectal cancer

Lingling Li; Yulu Hu; Yunshan Xu; Sanyi Tang

doi:10.1371/journal.pcbi.1011225

Abstract

Tumor heterogeneity is a large obstacle for cancer study and treatment. Different cancer patients may involve different combinations of gene mutations or the distinct regulatory pathways for inducing the progression of tumor. Investigating the pathways of gene mutations which can cause the formation of tumor can provide a basis for the personalized treatment of cancer. Studies suggested that KRAS, APC and TP53 are the most significant driver genes for colorectal cancer. However, it is still an open issue regarding the detailed mutation order of these genes in the development of colorectal cancer. For this purpose, we analyze the mathematical model considering all orders of mutations in oncogene, KRAS and tumor suppressor genes, APC and TP53, and fit it on data describing the incidence rates of colorectal cancer at different age from the Surveillance Epidemiology and End Results registry in the United States for the year 1973–2013. The specific orders that can induce the development of colorectal cancer are identified by the model fitting. The fitting results indicate that the mutation orders with KRAS → APC → TP53, APC → TP53 → KRAS and APC → KRAS → TP53 explain the age–specific risk of colorectal cancer with very well. Furthermore, eleven pathways of gene mutations can be accepted for the mutation order of genes with KRAS → APC → TP53, APC → TP53 → KRAS and APC → KRAS → TP53, and the alternation of APC acts as the initiating or promoting event in the colorectal cancer. The estimated mutation rates of cells in the different pathways demonstrate that genetic instability must exist in colorectal cancer with alterations of genes, KRAS, APC and TP53.

Author summary

Cumulative mutations in driver genes are the essential cause of cancer disease. For the colorectal cancer, KRAS, APC and TP53 are the common driver genes, and approximately 15% patients with colorectal cancer carry all mutations of the three genes. Exploring the pathway of mutations in these gene is extremely useful for the diagnosis and treatment of cancer. However, the mutation orders of these genes may vary in different patients due to the heterogeneity of tumor. Hence, we discuss all possible mutation orders in the genes, KRAS, APC and TP53 by using the model with five hits and find out the mutation pathways of genes that can effectively fit the incidence rate of colorectal cancer at different age in this article. In addition, we give the estimated values of mutation rates in each pathway that can explain the procession of colorectal cancer. The results obtained can offer guidelines to the treatment strategy of colorectal cancer.

Citation: Li L, Hu Y, Xu Y, Tang S (2023) Mathematical modeling the order of driver gene mutations in colorectal cancer. PLoS Comput Biol 19(6): e1011225. https://doi.org/10.1371/journal.pcbi.1011225

Editor: Philip K. Maini, Oxford, UNITED KINGDOM

Received: October 8, 2022; Accepted: May 29, 2023; Published: June 27, 2023

Copyright: © 2023 Li et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data cannot be shared publicly, since public availability would compromise patient privacy. The data can be accessed upon request https://www.seer.cancer.gov/seertrack/data/request/, which need complete the application form to obtain the SEER*Stat account.

Funding: The Natural Science Foundation of China supported financially this work: 12001417 to LL and 12031010 to ST. LL also received funding from Shaanxi Science and Technology Association Young Talent Lifting Program (grant number 20220519). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Colorectal cancer is one of the most common malignant tumors of digestive tract, which is the cancer with the third highest incidence and the second leading cause of mortality in the world [1]. Cancer is widely believed to have been caused by the accumulation of genetic and epigenetic alterations that result in the transformation of normal colonic epithelium to malignant cells. Recent studies showed that the progression from normal cells to the first malignant cell was supported by the alternations in three driver genes involving APC, KRAS and TP53, all of which were the most significant driver mutations in the colorectal cancer [2, 3]. However, not all colorectal cancers harbor the alterations of these three genes. Statistical analysis suggested that approximately 15% of colorectal cancers contained all mutations in APC, KRAS and TP53, and approximately 20% of tumor carried the mutations in both APC and KRAS [4, 5]. Among them, APC and TP53 are tumor suppressor genes (TSGs), and KRAS is an oncogene. It is well known that activation of oncogene, KRAS, only needs one hit, and inactivation of the TSGs, APC and TP53, requires two hits [6]. Hence, it takes five hits to turn the normal cells into the malignant cells for the primary colorectal cancer involving three driver genes, APC, KRAS and TP53.

Studying the mechanism of cancer formation by using the mathematical model is of essential importance in the prediction of tumor risk. The mathematical models considering the number of gene mutation were presented to explore the mechanism of cancerigenesis by a large number of scholars. The most revealing cancer models were the multistage model proposed by Armitage and Doll as early as 1950s [7, 8]. Their models were used to match age-specific mortality rate of various cancers, and the results indicated that the logarithm of mortality rate was linear with the logarithm of age. Subsequently, they considered the clonal expansion of cells and presented the two-stage model with clonal expansion of cells, widely applied to study the risk of various cancer [9]. Notably, Knudson utilized this model to fit incidence rate data of retinoblastoma and discovered the RB gene, the earliest TSG [10]. In addition, Moolgavkar et al. developed the method of solving the hazard function and the probability of tumor for the stochastic two-stage model involving selective growth advantage of cells and explained the detailed biological meaning of the model [11, 12]. However, two mutations are not sufficient to pose all cancers. Therefore, the two-stage model with clonal expansion of cells was extended to the model with more than two events where mutations accumulate in a specific order, allowing the intermediate stages to provide selective growth advantage to mutated cells [13–16]. It is showed that the model with more than two mutations better fits the incidence data of colorectal cancer compared to the two-stage model. Besides, Lang et al. used the two-type branching process model to study the dynamic behaviour of adenoma growth and transition into carcinomas in colorectal [17]. Nevertheless, these models did not consider the detailed mechanism of gene mutations.

The detailed mutation mechanism of driver gene is extremely urgent for colorectal cancer patients at different ages, which can provide insights into the strategies for detection and treatment of tumor with a more effective way. Recently, the mathematical model involving the specific driver genes of colorectal cancer was developed to analyze the procession of malignant transformation in the colorectal [18, 19]. The stochastic model of malignant transformation with the losses of APC and TP53 and gain of KRAS presented by Paterson et al. was used to study the specific order of these genes and indicated that inactivation of APC initiated the progression of tumor in more than half of colorectal cancer cases [18]. Subsequently, Zhang et al. studied the sizes of cells with different gene mutation and the waiting time distribution of driver gene mutations in colorectal cancer by using the five-hit branching process model based on the result of Paterson et al [20]. These work proposed the approximate analytic solution with only net growth of cells and analyzed the procession of malignant transformation in the colorectal cancer by using the solution and fixing the values of some parameters from references and some mutational data. However, they did not consider the effect of the death rate of cells on the risk of colorectal cancer. In addition, the rates of mutation and loss of heterozygosity (LOH) are not easy decided, since the rates of mutation and LOH may vary as the type of genetic instability and the mutation order of gene mutation in the colorectal cancer [6, 21]. Therefore, our work considers the effect of the growth rate and death rate of cells on the risk of colorectal cancer and does not employ the value of gene mutation rate gave by reference [18].

We use the mathematical model with five hits considering the selective advantage provided by the KRAS, APC and TP53 genes. In the model, any order of mutations in genes KRAS, APC and TP53 is allowed, leading to multiple evolutionary pathways that induce the development of colorectal cancer. Moreover, all possible mutation pathways of each order are considered. We first compare the approximate analytic solution and the exact numerical solution we used. Our result shows that approximate solution will get worse in estimating the risk of cancer at young ages (before 25 ages) when the growth rate or death rate of cells is enough big (e.g. 50 per year). Thus, the exact numerical solution of the model is chosen to explore the mechanism of mutations in genes KRAS, APC and TP53 that are most likely to occur in the development of colorectal cancer. In our fitting, approximate Bayesian computation schemes with the simulated likelihood density is used to estimate the mutation rates of model [22]. Finally, we analyze the models with specific mutation order that can fit well the specific-age incidence data of colorectal cancer and apply them to estimate the expected number of mutated cells with alternations of different genes in the early stages of colorectal carcinogenesis.

Materials and methods

Recent studies show that three driver genes are sufficient to produce the tumor in colorectal cancer, which involve two TSGs and one oncogene [3, 18, 23]. It is familiar that TSG needs two mutations to loss function keeping the growth of cells, and the oncogene is activated by a single mutation. Hence, the normal cells require five mutations to become tumor for colorectal cancer involving TP53, APC and KRAS. The detailed model is displayed in Fig 1.

Download:

Fig 1. The model with five mutations involving two TSGs (APC and TP53) and one oncogene (KRAS).

Pi (i = 1, 2, 3, 4) denotes the premalignant cells with i mutations, and M denotes the persistent malignant cells. μ_N denotes the mutation rate per normal stem cell per year, denotes the mutation rate per premalignant cell with i mutations per year, and denotes the net growth rate per premalignant cell per year, which is equal to be the difference between the rate of growth of cells and the death rate of cells. T_lag denotes the lag time from a persistent malignant cell to detectable tumor.

https://doi.org/10.1371/journal.pcbi.1011225.g001

In Fig 1, the net growth rate of cells, , is equal to the difference between growth rate of cells () and death rate of cells (). We define the following variables,

N—the number of normal cells in all crypts;
P_i(t)—the number of premalignant cells with i mutations at time t;
M(t)—the number of persistent malignant cells at time t;
D(t)—whether the tumor is detected clinically at time t, and the value is 1 if tumor detected, otherwise the value is 0.

In the model, represents the effective transformation rate from premalignant cell to persistent malignant one that does not die. Let τ be the time of the first persistent malignant cell. The persistent malignant cell is produced at rate at time s, the probability that persistent malignant cell doesn’t show up by time t yields [24], (1) Then the probability that a persistent malignant cell occurs by time t is given by (2)

Here, we are mainly concerned with the progression from normal cells to a persistent malignant cell, and the progression from a persistent malignant cell to the tumor detected clinically is assumed to be a fixed time, T_lag. Thus, the probability that tumor is clinically detected at time t can be written as (3) The hazard function, that is, the incidence rate of primary malignant tumors at time t, h(t), follows that (4) The expected number of premalignant cells can be decided by the following system of equations (5) The detailed derivation of the above equation is depicted in S1 Text. By solving Eqs (4) and (5), we get (6)

The above approximate solution is supported when the probability of persistent malignant cell, P(M(t) > 0), is close to zero [11]. However, it is not applied to cancers with high incidence rate [11]. Moreover, this approximate solution depends on the net growth rates and the mutation rates of cells, which doesn’t reflect the effects of the growth or death rate of cells on the risk of cancer. Therefore, we analyze another solution of hazard function. Here we consider the growth rate and death rate of cells in the development of cancer. Let and denote the growth rate and death rate of premalignant cells with i mutation(s), respectively.

To solve the hazard function, we define the following probability generating functions (7) It is easy to obtain the following survival function, that is, probability of no tumor at time t, (8) and then (9)

The probability generating functions in formulas (7) satisfy the following Kolmogorov backward equations [25, 26], (10) The detailed derivation of above equation can be seen in S1 Text. By the definitions of the probability generating functions, the initial values of equations (10) are as follows (11) Then, hazard function can be obtained by solving equations (10) with the initial conditions (11).

Results

Comparison of two solutions

To analyze the effect of the growth rate of cells on the risk of cancer, we fix the following parameter values and let the growth rate of cells, , takes different values, for example, , , and . The logarithm of hazard functions from formula (6) and equations (10) with the initial conditions (11) are illustrated in Fig 2(A), and the corresponding absolute errs are seen in Fig 2(C). From Fig 2(A) and 2(C), we find that the difference between the approximate solution and the exact numerical solution of the model goes up with the growth rate of cells increases, especially for older patients. The higher the growth rate , the larger the err between the approximate solution and the exact numerical solution of the model at an earlier time (younger ages). In like manner, the result holds true by varying death rate of cells (see Fig 2(B) and 2(D). For the pure birth process (), the difference between the approximate solution and the exact numerical solution of the model is still significant. Then, Fig 2 shows that the approximate solution is far greater than the exact numerical solution when the division rate or death rate of cells is quite large. In addition, the approximate solution only reflects the effect of net growth rate on hazard function. Nevertheless, the exact numerical solution involves the growth rate and death rate of cells in addition to net growth rate of cells. In fact, the approximate solution is given assuming that the probability of malignant cells is zero [11]. The approximate solution will get worse with an increase in the growth rate or death rate of cells. Therefore, the approximate solution may not be an excellent choice to simulate the risk of cancer, especially for a cancer with a high incidence rate.

Download:

Fig 2. Comparison of the approximate hazard function given by (6) with the exact numerical solution of hazard function given by (9).

All parameters of the model are set as follows: N = 10⁷, , λ_i = 0.1 + (i − 1) * 0.05.

https://doi.org/10.1371/journal.pcbi.1011225.g002

For stem cell in colon, it divides on average every five days [27]. It is shown that a stem cell produces two identical daughter cells by symmetry division and one mutated cell and one equivalent daughter cell by asymmetric division. In the model, the mutation rate, , signifies an asymmetric division rate of per cell per year. That is, the sum of the mutation rate () and the growth rate () of the model is equal to the division rate of per cell per year. However, the mutation rate, , is far less than the growth rate, . Therefore, it is easy to estimate the growth rate of stem cells, approximately 73 per cell per year. By the Fig 2, using the approximate solution may produce a large difference in estimating the risk of colorectal cancer compare to the exact numerical solution. As a consequence, we simulate the incidence rate of colorectal cancer at specific age by using the numerical solution obtained by equations (9), (10) and (11).

The order of gene alterations

Gene mutation is the leading cause of drug resistance whose emergence is a great obstacle for tumor treatment [28, 29]. For example, the mutation of KRAS leads to the resistant to gefitinib, erlotinib and cetuximab which induce the apoptosis of cancer cells in the therapy of cancer [30, 31]. Hence, knowing the order of gene alterations in a tumor is extremely crucial for the early treatment of cancer, which provides the guidance to the selection of medicines. Here, we investigate all possible mutation orders of three driver gene in the colorectal cancer by the mathematical modelling. Tumor, including all alterations of the three drive genes, involves six cases with the order of gene alterations. The detailed descriptions are seen in Fig 3. All possible pathways of alterations in the genes, KRAS, APC and TP53, are displayed in Figs A–F in S1 Text.

Download:

Fig 3. All orders of gene mutations for three driver genes, KRAS, APC and TP53, in colorectal cancer.

https://doi.org/10.1371/journal.pcbi.1011225.g003

TSG can inhibit tumor formation by inducing the death of abnormal cell, whose function is blocked only when both alleles are inactivated (TSG^−/−). In other words, the cells do not grow abnormally when only one allele is inactivated (TSG^+/−). However, the activation of oncogene just takes one hit (oncogene⁺), which promotes the proliferation of cells. Thus, by the clonal expansion of premalignant cells, all possible pathways for the alterations of three genes can be summarized as follows:

and , which involves the following sequences of mutations
- KRAS⁺ → APC^+/− → APC^−/− → TP53^+/− → TP53^−/−
- KRAS⁺ → TP53^+/− → TP53^−/− → APC^+/− → APC^−/−
, which involves the following sequences of mutations
- KRAS⁺ → APC^+/− → TP53^+/− → APC^−/− → TP53^−/−
- KRAS⁺ → TP53^+/− → APC^+/− → APC^−/− → TP53^−/−
- KRAS⁺ → APC^+/− → TP53^+/− → TP53^−/− → APC^−/−
- KRAS⁺ → TP53^+/− → APC^+/− → TP53^−/− → APC^−/−
and , which involves the following sequences of mutations
- APC^+/− → KRAS⁺ → TP53^+/− → APC^−/− → TP53^−/−
- TP53^+/− → KRAS⁺ → APC^+/− → APC^−/− → TP53^−/−
- APC^+/− → KRAS⁺ → TP53^+/− → TP53^−/− → APC^−/−
- TP53^+/− → KRAS⁺ → APC^+/− → TP53^−/− → APC^−/−
- APC^+/− → APC^−/− → TP53^+/− → TP53^−/− → KRAS⁺
- APC^+/− → APC^−/− → TP53^+/− → KRAS⁺ → TP53^−/−
- TP53^+/− → TP53^−/− → APC^+/− → APC^−/− → KRAS⁺
- TP53^+/− → TP53^−/− → APC^+/− → KRAS⁺ → APC^−/−
and , which involves the following sequences of mutations
- APC^+/− → KRAS⁺ → APC^−/− → TP53^+/− → TP53^−/−
- TP53^+/− → KRAS⁺ → TP53^−/− → APC^+/− → APC^−/−
- APC^+/− → APC^−/− → KRAS⁺ → TP53^+/− → TP53^−/−
- TP53^+/− → TP53^−/− → KRAS⁺ → APC^+/− → APC^−/−
and , which involves the following sequences of mutations
- APC^+/− → TP53^+/− → KRAS⁺ → APC^−/− → TP53^−/−
- TP53^+/− → APC^+/− → KRAS⁺ → APC^−/− → TP53^−/−
- APC^+/− → TP53^+/− → KRAS⁺ → TP53^−/− → APC^−/−
- TP53^+/− → APC^+/− → KRAS⁺ → TP53^−/− → APC^−/−
- APC^+/− → TP53^+/− → APC^−/− → TP53^−/− → KRAS⁺
- TP53^+/− → APC^+/− → APC^−/− → TP53^−/− → KRAS⁺
- APC^+/− → TP53^+/− → APC^−/− → KRAS⁺ → TP53^−/−
- TP53^+/− → APC^+/− → APC^−/− → KRAS⁺ → TP53^−/−
- APC^+/− → TP53^+/− → TP53^−/− → APC^−/− → KRAS⁺
- TP53^+/− → APC^+/− → TP53^−/− → APC^−/− → KRAS⁺
- APC^+/− → TP53^+/− → TP53^−/− → KRAS⁺ → APC^−/−
- TP53^+/− → APC^+/− → TP53^−/− → KRAS⁺ → APC^−/−

There are five cases for the model with five hits by the above analyses, involving 30 pathways of gene mutations. To determine the specific order of three drive genes alterations, we choose the numerical solution of the model to fit the incidence rate data of colorectal cancer from the Surveillance Epidemiology and End Results (SEER) registry during the period 1973–2013. However, not all parameters of the model can be estimated by the data alone. Here, we assumed the growth rate of normal cells to be 73 per year, and the lag time from a persistent malignant cell to tumor detected clinically to be 5 years [27, 32]. In addition, all mutation rates of cells () are limited in the range (0, 10⁻²), since the probability of a gene mutation ( t) is far less than one.

The activation of oncogene promotes the proliferation of cells, and the differentiation or apoptosis mechanism of cells will disarrange if the inactivation of TSG happens [6]. Studies showed that crypts carrying one of APC and KRAS alterations (that is the inactivation of APC and activation of KRAS) confer a selective growth advantage to cells [33, 34]. It has been shown that the net growth rate of the cells including only APC^−/− is 0.2 per year, and 0.07 per year for the cells with only KRAS⁺ [35, 36]. Furthermore, the inactivation of TP53 does not provide any growth advantage to cells in normal conditions [37]. We assume that mutations between genes have no interaction in cell growth. Therefore, the growth rate and death rate of premalignant cells with different gene alterations in the model are set to be as follows:

for the premalignant cells with TP53^−/−, the growth rate and death rate are 73 per year and 73 per year, respectively;
for the premalignant cells with KRAS⁺ or both TP53^−/− and KRAS⁺, the growth rate and death rate are 73.07 per year and 73 per year, respectively;
for the premalignant cells with APC^−/− or both TP53^−/− and APC^−/−, the growth rate and death rate are 73 per year and 72.8 per year, respectively;
for the premalignant cells with both APC^−/− and KRAS⁺, the growth rate and death rate are 73.07 per year and 72.8 per year, respectively.

In our simulations, the fourth-order Runge-Kutta is utilized to solve Equations (10) with the initial values (formulas (11)), and the model parameters, where v = Nμ_N, are estimated by using approximate Bayesian computation schemes involving the simulated likelihood density [22].

We do twenty fits of model with the different proliferation rate of cells (λ_pi), since mutation in first allele of TP53 and APC (TP53^+/− and APC^+/−) does not bring about the change in proliferation rate of cell. These fits correspond to the pathways in Figs A–F in S1 Text. Our fitting results suggest that only the orders with KRAS−APC−TP53, APC−TP53−KRAS and APC−KRAS−TP53 can accepted to explain the incidence rate of colorectal cancer at different ages, involving eleven pathways (see Fig 4). There are four pathway of mutations for the orders with KRAS − APC − TP53 and APC − KRAS − TP53, and three sequences of mutations for the order with APC − TP53 − KRAS. As a consequence, the first event is just activation of KRAS or inactivation of APC, and the inactivation of TP53 is usually a late event in the development of colorectal cancer.

Download:

Fig 4. All pathways are accepted to explain the development of colorectal cancer at different ages.

https://doi.org/10.1371/journal.pcbi.1011225.g004

By the previous analyses, the five–stage models with KRAS⁺ → APC^+/− → TP53^+/− → APC^−/− → TP53^−/− and KRAS⁺ → TP53^+/− → APC^+/− → APC^−/− → TP53^−/− (correspond to Fig A3 of Fig A in S1 Text) have the same parameter values for the order with KRAS − APC − TP53. It is the same for the order with APC − TP53 − KRAS involving APC^+/− → TP53^+/− → APC^−/− → TP53^−/− → KRAS⁺ and TP53^+/− → APC^+/− → APC^−/− → TP53^−/− → KRAS⁺ (correspond to Fig C2 of Fig C in S1 Text), and the order with APC − KRAS − TP53 involving APC^+/− → TP53^+/− → APC^−/− → KRAS⁺ → TP53^−/− and TP53^+/− → APC^+/− → APC^−/− → KRAS⁺ → TP53^−/− (correspond to Fig D3 of Fig D in S1 Text). The fitting results of the pathways in Fig 4 are displayed in Figs 5–12, and the detailed values of parameters are shown in Tables 1–8. They show that the first mutation rate is extremely low for all pathways in Fig 4, and the mutation rate of the same gene is different in different mutation pathways. These parameter estimates can provide some evidences for inferring the type of gene alteration.

Download:

Fig 5. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with KRAS⁺ → APC^+/− → APC^−/− → TP53^+/− → TP53^−/−.

(a) (Nμ_N of model), (b) ( of model), (c) ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g005

Download:

Fig 6. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with APC^+/− → KRAS⁺ → APC^−/− → TP53^+/− → TP53^−/−.

(a) (Nμ_N of model), (b) ( of model), (c) ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g006

Download:

Fig 7. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with KRAS⁺ → APC^+/− → TP53^+/− → APC^−/− → TP53^−/− or KRAS⁺ → TP53^+/−→ APC^+/− →APC^−/− →TP53^−/−.

(a) (Nμ_N of model), (b) or ( of model), (c) or ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g007

Download:

Fig 8. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with APC^+/− → APC^−/− → TP53^+/− → TP53^−/− → KRAS⁺.

(a) (Nμ_N of model), (b) ( of model), (c) ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g008

Download:

Fig 9. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with APC^+/− → TP53^+/− → APC^−/− → TP53^−/− → KRAS⁺ or TP53^+/− → APC^+/− → APC^−/− → TP53^−/− → KRAS⁺.

(a) or (Nμ_N of model), (b) or ( of model), (c) ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g009

Download:

Fig 10. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with APC^+/− → APC^−/− → KRAS⁺ → TP53^+/− → TP53^−/−.

(a) (Nμ_N of model), (b) ( of model), (c) ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g010

Download:

Fig 11. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with APC^+/− → APC^−/− → TP53^+/− → KRAS⁺ → TP53^−/−.

(a) (Nμ_N of model), (b) ( of model), (c) ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g011

Download:

Fig 12. The fitting result and Probabilistic distributions of the estimated parameters using approximate Bayesian computation schemes involving the simulated likelihood density for the pathway with APC^+/− → TP53^+/− → APC^−/− → KRAS⁺ → TP53^−/− or TP53^+/− → APC^+/− → APC^−/− → KRAS⁺ → TP53^−/−.

(a) or (Nμ_N of model), (b) or ( of model), (c) ( of model), (d) ( of model), (e) ( of model), (f) The colorectal cancer incidence rate per 100,000 patients from SEER registry and rates predicted by the model.

https://doi.org/10.1371/journal.pcbi.1011225.g012

Download:

Table 1. Estimate values of parameters in the pathway with KRAS⁺ → APC^+/− → APC^−/− → TP53^+/− → TP53^−/−.

https://doi.org/10.1371/journal.pcbi.1011225.t001

Download:

Table 2. Estimate values of parameters in the pathway with APC^+/− → KRAS⁺ → APC^−/− → TP53^+/− → TP53^−/−.

https://doi.org/10.1371/journal.pcbi.1011225.t002

Download:

Table 3. Estimate values of parameters in the pathway with KRAS⁺ → APC^+/− → TP53^+/− → APC^−/− → TP53^−/− or KRAS⁺ → TP53^+/− → APC^+/− → APC^−/− → TP53^−/−.

https://doi.org/10.1371/journal.pcbi.1011225.t003

Download:

Table 4. Estimate values of parameters in the pathway with APC^+/− → APC^−/− → TP53^+/− → TP53^−/− → KRAS⁺.

https://doi.org/10.1371/journal.pcbi.1011225.t004

Download:

Table 5. Estimate values of parameters in the pathway with APC^+/− → TP53^+/− → APC^−/− → TP53^−/− → KRAS⁺ or TP53^+/− → APC^+/− → APC^−/− → TP53^−/− → KRAS⁺.

https://doi.org/10.1371/journal.pcbi.1011225.t005

Download:

Table 6. Estimate values of parameters in the pathway with APC^+/− → APC^−/− → KRAS⁺ → TP53^+/− → TP53^−/−.

https://doi.org/10.1371/journal.pcbi.1011225.t006

Download:

Table 7. Estimate values of parameters in the pathway with APC^+/− → APC^−/− → TP53^+/− → KRAS⁺ → TP53^−/−.

https://doi.org/10.1371/journal.pcbi.1011225.t007

Download:

Table 8. Estimate values of parameters in the pathway with APC^+/− → TP53^+/− → APC^−/− → KRAS⁺ → TP53^−/− or TP53^+/− → APC^+/− → APC^−/− → KRAS⁺ → TP53^−/−.

https://doi.org/10.1371/journal.pcbi.1011225.t008

For the pathway with KRAS⁺ → APC^+/− → APC^−/− → TP53^+/− → TP53^−/−, Table 1 shows the rate of mutation on first APC allele is larger than that of mutation on second APC allele, approximately 2 times, which infers that inactivation of APC is caused by mutation in both alleles, or by LOH in first allele and mutation in second allele. In addition, the inactivation rates of APC and TP53 are much greater than the activation rate of KRAS and the point mutation rate [38, 39]. The total number of driver positions in APC is 604 from reference [18], and then the mutation rate of APC is 7.55 × 10⁻⁶ if the base pair mutation rate takes 1.25 × 10⁻⁸ from [39]. Our estimated value of alteration in APC (4.70 × 10⁻³) is much larger than 7.55 × 10⁻⁶. These imply that the inactivations of APC or TP53 may be accompanied by genetic instability (microsatellite instability or chromosomal instability) that increases the mutation rate of gene. Evidences manifested that chromosomal instability might result from mutations in APC or TP53 [40, 41].

By comparing the mutation rate of the first event in the orders with KRAS − APC − TP53, APC − TP53 − KRAS and APC − KRAS − TP53, we find that the initiated rate in the order with KRAS − APC − TP53 is lower than that in other orders with high probability in initiation of colorectal cancer. Tables 1–8 indicate that the transform rate (μ_p4) of malignant cells is an extraordinarily high value for all pathways. It is shown that genetic instability must exist in the development of colorectal cancer. In addition, inactivation rate of APC () in the pathways with APC^+/− → APC^−/− → TP53^+/− → TP53^−/− → KRAS⁺, APC^+/− → APC^−/− → KRAS⁺ → TP53^+/− → TP53^−/− and APC^+/− → APC^−/− → TP53^+/− →KRAS⁺ → TP53^−/− is much less than that in other pathways, approximately . Evidences suggested that the point mutation rate of genes with microsatellite instability is about ten times as much as that without microsatellite instability for colorectal cancer [42, 43]. It can be used as a mark in identifying the detailed mutation pathway for the orders with APC − TP53 − KRAS and APC − KRAS − TP53.

The number of cells with different genetic alterations

We next use the five-stage models with KRAS − APC − TP53, APC − TP53 − KRAS and APC − KRAS − TP53 to detect the changes of the mutated cells with different genetic mutations over time. By Fig 4, the cells with single genetic mutation have two types: the cells with activation of KRAS and those with inactivation of APC. In addition, there are three types of mutated cells with the combination of two genetic mutations. They are mutated cells with activation of KRAS followed by inactivation of APC, those with inactivation of APC followed by activation of KRAS, and those with inactivation of APC followed by inactivation of TP53. The detailed formulas about the expected numbers of mutated cells with different genetic alterations are given in S1 Text, and the changes in the number of mutated cells over time in different order of gene alterations are plotted in Fig 13.

Fig 13 shows that the number of KRAS-mutated cells is much higher than that of cells with both activated KRAS and inactivated APC at the beginning, and then the number of cells with both KRAS-mutated and APC-mutated surpasses that of KRAS-mutated cells by middle age in the order with KRAS − APC − TP53. For the orders with APC − TP53 − KRAS and APC − KRAS − TP53, the number of cells with single-mutant outnumber those with double-mutant, throughout the human lifetime. Additionally, the number of cells with double-mutant is close to those with a single gene mutation over time for the order with APC − KRAS − TP53. These changes in the number of cells with different gene alterations are able to offer some clues in identifying the detailed order of gene alterations for patients with colorectal cancer.

Download:

Fig 13. The logarithm of expected number of mutated cells with the combination of different genetic mutations for the pathways involving KRAS⁺ → APC^+/− → APC^−/− → TP53^+/− → TP53^−/−, APC^+/− → APC^−/− → TP53^+/− → TP53^−/− → KRAS⁺ and APC^+/− → APC^−/− → KRAS⁺ → TP53^+/− → TP53^−/−.

The values of parameters are decided by Tables 1–8.

https://doi.org/10.1371/journal.pcbi.1011225.g013

Discussion

The order of driver gene mutations is very vital for the clinical treatment and even prevention in the course of cancer. In this article, we construct the mathematical model considering all possible orders of mutations in APC, TP53 and KRAS that are the most significant driver genes in colorectal cancer. There are 30 pathways involving the alterations of all three driver genes, which need five hits from a normal cell to a malignant cell with all mutations of these driver genes. These pathways are classified as twenty cases based on the different proliferation rates of mutated cells, which corresponds with the five-stage model with different net growth rate of mutated cells (see Figs A–F in S1 Text). Studies showed that approximately 15% colorectal cancer harbored alterations in all of the three genes [4, 5]. Therefore, we use the constructed model based on the net proliferation rates of mutated cells to match 15% incidence rate of colorectal cancer for male and female in different ages instead of those of a fixed age [18]. Our fitting results indicate that three mutation orders with KRAS → APC → TP53, APC → TP53 → KRAS, APC → KRAS → TP53, are supported to lead to tumor, involving eleven pathways. Among them, there are seven pathways that inactivation of APC acts as the first event in the colorectal cancer, and the inactivation of TP53 is the last event when the first event is activation of KRAS in the development of colorectal cancer. This is line with the results that the alternations of APC is an early driver event that accounts for 80% of colorectal cancers [44, 45]. Inactivation of TP53 is critical in the transformation from early adenoma to advanced tumor, which regulates G1 cycle and apoptosis of cells. As a result, the inactivation of TP53 is usually a late event in the development of colorectal cancer. drugs that do not develop the resistance caused by the mutation of KRAS and APC should be adopted preferentially in the therapy of colorectal cancer.

The multi–type branching process is a useful tool to solve the risk function of the multistage model. The approximate solution of hazard function is used to predict the risk of cancer by some work [18, 46, 47]. Here, we consider the effect of the growth rate and death rate of cells on risk cancer and make a comparison for the approximate solution without the death rate of cells and the exact numerical solution including the growth rate and death rate of cells. It is shown that the approximate solution will overestimate the risk of cancer compare to the exact numerical solution when all parameters of model are fixed, especially for high growth rate or death rate of cells. As a consequence, we choose the exact numerical solution of hazard function to fit the data with incidence rate of colorectal cancer from SEER register during the year 1973–2013. In our work, we do not fix all parameters of model like references [18], and only growth rate and death rate of cells are decided from published references. The mutation rates of gene are not accurately inferred only based on the mutational data due to genetic instability and the type of gene mutation. For this purpose, the mutation rates of model are estimated by fitting incidence rate data of colorectal cancer based on approximate Bayesian computation schemes with simulated likelihood density. We consider the simulation of incidence rate data of colorectal cancer from 0 to 84 years old instead of that at 80 years old in reference [18]. The estimated mutation rates are extremely valuable to predict the risk of colorectal cancer. Moreover, approximate Bayesian computation schemes is a good method to estimate the biological parameters. Among them, approximate Bayesian computation with Markov chain Monte Carlo and sequential Monte Carlo was applied to biological systems with success [48, 49]. Here, we use approximate Bayesian computation schemes with simulated likelihood density to estimate our model parameters, which is effective for inferring parameters in high-dimensional stochastic model, and can reduce the probability of getting stuck the low probability regions compared to approximate Bayesian computation with Markov chain Monte Carlo [22].

To further identify the feature of different gene mutation orders, the sizes of premalignant cells in the evolutionary pathways with KRAS → APC → TP53, APC → TP53 → KRAS, APC → KRAS → TP53 are analyzed. We find that the sizes of cells with single-mutant and double-mutant are significant difference in different gene mutation orders. The number of KRAS–mutated cells is higher than that of cells with both activated KRAS and inactivated APC at the beginning, and the number of cells with both KRAS–activated and APC–inactivated will surpass that of KRAS–mutated cells over time for the order with KRAS → APC → TP53. For the orders with APC → TP53 → KRAS and APC → KRAS → TP53, the cells with APC^−/− far outnumber those with APC^−/−TP53^−/− and APC^−/−KRAS⁺. These results can be used as the mark of colorectal cancer diagnosis and inferring the sequence of gene mutation.

In this paper, we not only give the mutation order of three driver genes, but also provide the detailed pathway of gene mutations. However, there are still some limits in our work. Our results obtained do not consider the interaction effect between mutations in two genes, since no evidence suggests that epistatic interactions occur in mutations of two genes [4]. As is well known, inactivation of each allele of TSG has two ways involving point mutation and the LOH. We do not discuss the detailed pattern of alteration in TSGs, APC and TP53 in our model. The reason is that the model involving the pattern of inactivation in TSGs includes too many parameters, which results in non–identifiability issue of model parameters. Moreover, the type of genomic instability is unclear because of heterogeneity of tumor. However, we discuss the mutation rate of gene in different pathway obtained by fitting the incidence rate data of colorectal cancer at different ages. These estimated values of mutation rates can provide some information for inferring the pattern of alteration in TSGs and the existence of genetic instability. With the development of DNA sequencing, sequenced cancer genomes and gene expression data are important data source for inferring the evolutionary pathway in cancer progression, and some methods are developed to predict tumor evolution [50–52]. Although these data and methods do not give the rate of gene mutation, they can provide new ideas for constructing specific mathematical model of cancer. Thus, more diverse data deserve to be considered in further studying evolutionary pathways of cancer, which will be the direction for future research.

Supporting information

S1 Text. Mathematical derivation and supplementary figures.

Fig A: All pathways of gene mutations for the case (A) in colorectal cancer. Fig B: All pathways of gene mutations for the case (B) in colorectal cancer. Fig C: All pathways of gene mutations for the case (C) in colorectal cancer. Fig D: All pathways of gene mutations for the case (D) in colorectal cancer. Fig E: All pathways of gene mutations for the case (E) in colorectal cancer. Fig F: All pathways of gene mutations for the case (F) in colorectal cancer.

https://doi.org/10.1371/journal.pcbi.1011225.s001

(PDF)

Acknowledgments

We thank Dr. Tianhai Tian for the help of code in the manuscript.

References

1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA. Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6): 394–424. pmid:30207593
- View Article
- PubMed/NCBI
- Google Scholar
2. Fearon ER. Molecular genetics of colorectal cancer. Annu Rev Pathol. 2011;6: 479–507. pmid:21090969
- View Article
- PubMed/NCBI
- Google Scholar
3. Tomasetti C, Marchionni L, Nowak MA, Parmigiani G, Vogelstein B. Only three driver gene mutations are required for the development of lung and colorectal cancers. Proc Natl Acad Sci U S A. 2015;112(1): 118–123. pmid:25535351
- View Article
- PubMed/NCBI
- Google Scholar
4. Lawrence MS, Stojanov P, Mermel CH, Robinson JT, Garraway LA, Golub TR, et al. Discovery and saturation analysis of cancer genes across 21 tumour types. Nature. 2014;505(7484): 495–501. pmid:24390350
- View Article
- PubMed/NCBI
- Google Scholar
5. Bos JL, Fearon ER, Hamilton SR, Vries MV, van Boom JH, van der Eb AJ, et al. Prevalence of ras gene-mutations in human colorectal cancers. Nature. 1987;327(6120): 293–297. pmid:3587348
- View Article
- PubMed/NCBI
- Google Scholar
6. Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med. 2004;10(8): 789–799. pmid:15286780
- View Article
- PubMed/NCBI
- Google Scholar
7. Nordling CO. A new theory of cancer inducing mechanism. Br J Cancer. 1953;7(1): 68–72. pmid:13051507
- View Article
- PubMed/NCBI
- Google Scholar
8. Armitage P, Doll R. The age distribution of cancer and multi–stage theory of carcinogenesis. Br J Cancer. 1954;8(1): 1–12. pmid:13172380
- View Article
- PubMed/NCBI
- Google Scholar
9. Armitage P, Doll R. A two-stage theory of carcinogenesis in relation to the age distribution of human cancer. Br J Cancer. 1957;9(12): 161–169.
- View Article
- Google Scholar
10. Knudson AG. Mutation and cancer: Statistical study of retinoblastoma. Proc Natl Acad Sci U S A. 1971;68(4): 820–823. pmid:5279523
- View Article
- PubMed/NCBI
- Google Scholar
11. Moolgavkar SH, Dewanji A, Venzon DJ. A stochastic two–stage model for cancer risk assessment I: The hazard function and the probability of tumor. Risk Anal. 1988;8(3): 383–392. pmid:3201016
- View Article
- PubMed/NCBI
- Google Scholar
12. Meza R, Jeon J, Moolgavkar SH, Luebeck EG. Age–specific incidence of cancer: phases, transitions, and biological implications. Proc Natl Acad Sci U S A. 2008;105(42): 16284–16289. pmid:18936480
- View Article
- PubMed/NCBI
- Google Scholar
13. Moolgavkar SH, Luebeck EG. Multistage carcinogenesis: population–based model for colon cancer. J Nat Cancer Inst. 1992;84(8): 610–618. pmid:1313509
- View Article
- PubMed/NCBI
- Google Scholar
14. Luebeck EG, Moolgavkar SH. Multistage carcinogenesis and the incidence of colorectal cancer. Proc Natl Acad Sci U S A. 2002;99(23): 15095–15100. pmid:12415112
- View Article
- PubMed/NCBI
- Google Scholar
15. Little MP, Wright EG. A stochastic carcinogenesis model incorporating genemic instability fitted to colon cancer data. Math Biosci. 2003;183(2): 111–134. pmid:12711407
- View Article
- PubMed/NCBI
- Google Scholar
16. Little MP, Vineis P, Li G. A stochastic carcinogenesis model incorporating multiple types of genomic instability fitted to colon cancer data. J Theor Biol. 2017;254(2): 229–238.
- View Article
- Google Scholar
17. Lang BM, Kuipers J, Misselwitz B, Beerenwinkel N. Predicting colorectal cancer risk from adenoma detection via a two-type branching process model. PLoS Comput Biol. 2020;16(2): e1007552. pmid:32023238
- View Article
- PubMed/NCBI
- Google Scholar
18. Paterson C, Clevers H, Bozic I. Mathematical model of colorectal cancer initiation. Proc Natl Acad Sci U S A. 2020;117(34): 20681–20688. pmid:32788368
- View Article
- PubMed/NCBI
- Google Scholar
19. Haupt S, Zeilmann A, Ahadova A, Blker H, Heuveline V. Mathematical modeling of multiple pathways in colorectal carcinogenesis using dynamical systems with kronecker structure. PLoS Comput Biol. 2021;17(5): e1008970. pmid:34003820
- View Article
- PubMed/NCBI
- Google Scholar
20. Zhang R, Ukogu OA, Bozic I. Waiting times in a branching process model of colorectal cancer initiation. Theor Popul Biol. 2023;151: 44–63. pmid:37100121
- View Article
- PubMed/NCBI
- Google Scholar
21. Lengauer C, Kinzler KW, Vogelstein B. Genetic instabilities in human cancers. Nature. 1998;396(6712): 643–649. pmid:9872311
- View Article
- PubMed/NCBI
- Google Scholar
22. Wu Q, Smith-Miles K, Tian T. Approximate Bayesian computation schemes for parameter inference of discrete stochastic models using simulated likelihood density. BMC Bioinform. 2014; 15(Suppl 12): S3. pmid:25473744
- View Article
- PubMed/NCBI
- Google Scholar
23. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA Jr, Kinzler KW. Cancer genome landscapes. Science. 2013;339(6127): 1546–1558. pmid:23539594
- View Article
- PubMed/NCBI
- Google Scholar
24. Durrett R. Branching Process Models of Cancer. In: Branching Process Models of Cancer. Mathematical Biosciences Institute Lecture Series, vol 1.1. Springer, Cham; 2015.
25. Harris TE. The theory of branching processes. 1st ed. Springer-Verlag Berlin Heidelberg; 1963.
26. Portier CJ, Sherman CKA. Calculating tumor incidence rates in stochastic models of carcinogenesis. Math Biosci. 1996;135(2): 129–146. pmid:8768218
- View Article
- PubMed/NCBI
- Google Scholar
27. Tomasetti C, Vogelstein B. Cancer etiology. Variation in cancer risk among tissues can be explained by the number of stem cell divisions. Science. 2015;347(6217): 78–81. pmid:25554788
- View Article
- PubMed/NCBI
- Google Scholar
28. Komarova NL, Wodarz D. Drug resistance in cancer: principles of emergence and prevention. Proc Natl Acad Sci U S A. 2005;102(27): 9714–9719. pmid:15980154
- View Article
- PubMed/NCBI
- Google Scholar
29. Shamieh SE, Saleh F, Assaad S, Farhat F. Next-generation sequencing reveals mutations in RB1, CDK4 and TP53 that may promote chemo-resistance to palbociclib in ovarian cancer. Drug Metabol Therapy. 2019;34(2): 20180027. pmid:31145688
- View Article
- PubMed/NCBI
- Google Scholar
30. Lièvre A, Bachet JB, Corre DL, Boige V, Landi B, Emile JF, et al. KRAS mutation status is predictive of response to cetuximab therapy in colorectal cancer. Cancer Res. 2012;66(8): 3992–3995.
- View Article
- Google Scholar
31. Pao W, Wang TY, Riely GJ, Miller VA, Pan Q, Ladanyi M, et al. KRAS mutations and primary resistance of lung adenocarcinomas to gefitinib or erlotinib. PLoS Medicine. 2005;2(1): e17. pmid:15696205
- View Article
- PubMed/NCBI
- Google Scholar
32. Luebeck EG, Curtius K. Impact of tumor progression on cancer incidence curves. Cancer Res. 2013;73(3): 1086–1096. pmid:23054397
- View Article
- PubMed/NCBI
- Google Scholar
33. Lamlum H, Papadopoulou A, Ilyas M, Rowan A, Gillet C, Hanby A, et al. APC mutations are sufficient for the growth of early colorectal adenomas. Proc Natl Acad Sci U S A. 2000;97(5): 2225–2228. pmid:10681434
- View Article
- PubMed/NCBI
- Google Scholar
34. Snippert HJ, Schepers AG, van Es JH, Simons BD, Clevers H. Biased competition between Lgr5 intestinal stem cells driven by oncogenic mutation induces clonal expansion. EMBO Rep. 2014;15(1): 62–69. pmid:24355609
- View Article
- PubMed/NCBI
- Google Scholar
35. Baker AM, Cereser B, Melton S, Fletcher AG, Rodriguez-Justo M, Tadrous PJ, et al. Quantification of crypt and stem cell evolution in the normal and neoplastic human colon. Cell Rep. 2014;8(4): 940–947. pmid:25127143
- View Article
- PubMed/NCBI
- Google Scholar
36. Nicholson AM, Olpe C, Hoyle A, Thorsen AS, Rus T, Colombé M, et al. Fixation and spread of somatic mutations in adult human colonic epithelium. Cell Stem Cell. 2018;22(6): 909–918. pmid:29779891
- View Article
- PubMed/NCBI
- Google Scholar
37. Vermeulen L, Morrissey E, van der Heijden D, Nicholson AM, Sottoriva A, Buczacki S, et al. Defining stem cell dynamics in models of intestinal tumor initiation. Science. 2013;342(6161): 995–998. pmid:24264992
- View Article
- PubMed/NCBI
- Google Scholar
38. Tomasetti C, Bozic I. The (not so) immortal strand hypothesis. Stem Cell Res. (Amst.) 2015;14(2): 238–241. pmid:25700960
- View Article
- PubMed/NCBI
- Google Scholar
39. Blokzijl F, Ligt JD, Jager M, Sasselli V, Roerink S, Sasaki N, et al. Tissue-specific mutation accumulation in human adult stem cells during life. Nature. 2016;538(7624): 260–264. pmid:27698416
- View Article
- PubMed/NCBI
- Google Scholar
40. Fukasawa K, Choi T, Kuriyama R, Rulong S, Vande Woude GF. Abnormal centrosome amplification in the absence of p53. Science. 1996;271(5256): 1744–1747. pmid:8596939
- View Article
- PubMed/NCBI
- Google Scholar
41. Fodde R, Kuipers J, Rosenberg C, Smits R, Kielman M, Gaspar C, et al. Mutations in the APC tumour suppressor gene cause chromosomal instability. Nature Cell Biol. 2001;3(4): 433–438. pmid:11283620
- View Article
- PubMed/NCBI
- Google Scholar
42. Timmermann B, Kerick M, Roehr C, Fischer A, Isau M, Boerno ST, et al. Somatic mutation profiles of MSI and MSS colorectal cancer identified by whole exome next generation sequencing and bioinformatics analysis. PLoS One. 2010;5(12): e15661. pmid:21203531
- View Article
- PubMed/NCBI
- Google Scholar
43. Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature. 2012;487(7407): 330–337.
- View Article
- Google Scholar
44. Powell SM, Zilz N, Beazer-Barclay Y, Bryan TM, Hamilton SR, Thibodeau SN, et al. APC mutations occur early during colorectal tumorigenesis. Nature. 1992;359(6392): 235–237. pmid:1528264
- View Article
- PubMed/NCBI
- Google Scholar
45. Suraweera N, Duval A, Reperant M, Vaury C, Furlan D, Leroy K, et al. Evaluation of tumor microsatellite instability using five quasimonomorphic repeats and pentaplex PCR. Gastroenterology. 2002;123(6): 1804–1811.
- View Article
- Google Scholar
46. Zhang X, Fang Y, Zhao YD, Zheng WM. Mathematical modeling the pathway of human breast cancer. Math Biosci. 2014;253: 25–29. pmid:24680645
- View Article
- PubMed/NCBI
- Google Scholar
47. Paterson C, Bozic I, Smith MJ, Hoad X, Evans DGR. A mechanistic mathematical model of initiation and malignant transformation in sporadic vestibular schwannoma. Br J Cancer. 2022;127: 1843–1857. pmid:36097176
- View Article
- PubMed/NCBI
- Google Scholar
48. Golightly A, Wilkinson DJ. Bayesian parameter inference for stochastic biochemical network models using particle Markov chain Monte Carlo. Interface Focus. 2011;1(6): 807–820. pmid:23226583
- View Article
- PubMed/NCBI
- Google Scholar
49. Bollen Y, Stelloo E, van Leenen P, van den Bos M, Ponsioen B, Lu BX, et al. Reconstructing single-cell karyotype alterations in colorectal cancer identifies punctuated and gradual diversification patterns. Nat Genet. 2021;53(8): 1187–1195. pmid:34211178
- View Article
- PubMed/NCBI
- Google Scholar
50. Caravagna G, Graudenzi A, Ramazzotti D, Sanz-Pamplona R, Sano LD, Mauri G, et al. Algorithmic methods to infer the evolutionary trajectories in cancer progression. Proc Natl Acad Sci U S A. 2016;113(28): E4025–E4034 pmid:27357673
- View Article
- PubMed/NCBI
- Google Scholar
51. Fleck JL, Pavel AB, Cassandras CG. Integrating mutation and gene expression cross-sectional data to infer cancer progression. BMC Syst Biol. 2016;10(1): 1–12. pmid:26810975
- View Article
- PubMed/NCBI
- Google Scholar
52. Diaz-Colunga J, Diaz-Uriarte R. Conditional prediction of consecutive tumor evolution using cancer progression models: What genotype comes next? PLoS Comput Biol. 2021;17(12): e1009055–e1009055. pmid:34932572
- View Article
- PubMed/NCBI
- Google Scholar

[ref1] 1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA. Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6): 394–424. pmid:30207593
View Article
PubMed/NCBI
Google Scholar

[2] View Article

[3] PubMed/NCBI

[4] Google Scholar

[ref2] 2. Fearon ER. Molecular genetics of colorectal cancer. Annu Rev Pathol. 2011;6: 479–507. pmid:21090969
View Article
PubMed/NCBI
Google Scholar

[6] View Article

[7] PubMed/NCBI

[8] Google Scholar

[ref3] 3. Tomasetti C, Marchionni L, Nowak MA, Parmigiani G, Vogelstein B. Only three driver gene mutations are required for the development of lung and colorectal cancers. Proc Natl Acad Sci U S A. 2015;112(1): 118–123. pmid:25535351
View Article
PubMed/NCBI
Google Scholar

[10] View Article

[11] PubMed/NCBI

[12] Google Scholar

[ref4] 4. Lawrence MS, Stojanov P, Mermel CH, Robinson JT, Garraway LA, Golub TR, et al. Discovery and saturation analysis of cancer genes across 21 tumour types. Nature. 2014;505(7484): 495–501. pmid:24390350
View Article
PubMed/NCBI
Google Scholar

[14] View Article

[15] PubMed/NCBI

[16] Google Scholar

[ref5] 5. Bos JL, Fearon ER, Hamilton SR, Vries MV, van Boom JH, van der Eb AJ, et al. Prevalence of ras gene-mutations in human colorectal cancers. Nature. 1987;327(6120): 293–297. pmid:3587348
View Article
PubMed/NCBI
Google Scholar

[18] View Article

[19] PubMed/NCBI

[20] Google Scholar

[ref6] 6. Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med. 2004;10(8): 789–799. pmid:15286780
View Article
PubMed/NCBI
Google Scholar

[22] View Article

[23] PubMed/NCBI

[24] Google Scholar

[ref7] 7. Nordling CO. A new theory of cancer inducing mechanism. Br J Cancer. 1953;7(1): 68–72. pmid:13051507
View Article
PubMed/NCBI
Google Scholar

[26] View Article

[27] PubMed/NCBI

[28] Google Scholar

[ref8] 8. Armitage P, Doll R. The age distribution of cancer and multi–stage theory of carcinogenesis. Br J Cancer. 1954;8(1): 1–12. pmid:13172380
View Article
PubMed/NCBI
Google Scholar

[30] View Article

[31] PubMed/NCBI

[32] Google Scholar

[ref9] 9. Armitage P, Doll R. A two-stage theory of carcinogenesis in relation to the age distribution of human cancer. Br J Cancer. 1957;9(12): 161–169.
View Article
Google Scholar

[34] View Article

[35] Google Scholar

[ref10] 10. Knudson AG. Mutation and cancer: Statistical study of retinoblastoma. Proc Natl Acad Sci U S A. 1971;68(4): 820–823. pmid:5279523
View Article
PubMed/NCBI
Google Scholar

[37] View Article

[38] PubMed/NCBI

[39] Google Scholar

[ref11] 11. Moolgavkar SH, Dewanji A, Venzon DJ. A stochastic two–stage model for cancer risk assessment I: The hazard function and the probability of tumor. Risk Anal. 1988;8(3): 383–392. pmid:3201016
View Article
PubMed/NCBI
Google Scholar

[41] View Article

[42] PubMed/NCBI

[43] Google Scholar

[ref12] 12. Meza R, Jeon J, Moolgavkar SH, Luebeck EG. Age–specific incidence of cancer: phases, transitions, and biological implications. Proc Natl Acad Sci U S A. 2008;105(42): 16284–16289. pmid:18936480
View Article
PubMed/NCBI
Google Scholar

[45] View Article

[46] PubMed/NCBI

[47] Google Scholar

[ref13] 13. Moolgavkar SH, Luebeck EG. Multistage carcinogenesis: population–based model for colon cancer. J Nat Cancer Inst. 1992;84(8): 610–618. pmid:1313509
View Article
PubMed/NCBI
Google Scholar

[49] View Article

[50] PubMed/NCBI

[51] Google Scholar

[ref14] 14. Luebeck EG, Moolgavkar SH. Multistage carcinogenesis and the incidence of colorectal cancer. Proc Natl Acad Sci U S A. 2002;99(23): 15095–15100. pmid:12415112
View Article
PubMed/NCBI
Google Scholar

[53] View Article

[54] PubMed/NCBI

[55] Google Scholar

[ref15] 15. Little MP, Wright EG. A stochastic carcinogenesis model incorporating genemic instability fitted to colon cancer data. Math Biosci. 2003;183(2): 111–134. pmid:12711407
View Article
PubMed/NCBI
Google Scholar

[57] View Article

[58] PubMed/NCBI

[59] Google Scholar

[ref16] 16. Little MP, Vineis P, Li G. A stochastic carcinogenesis model incorporating multiple types of genomic instability fitted to colon cancer data. J Theor Biol. 2017;254(2): 229–238.
View Article
Google Scholar

[61] View Article

[62] Google Scholar

[ref17] 17. Lang BM, Kuipers J, Misselwitz B, Beerenwinkel N. Predicting colorectal cancer risk from adenoma detection via a two-type branching process model. PLoS Comput Biol. 2020;16(2): e1007552. pmid:32023238
View Article
PubMed/NCBI
Google Scholar

[64] View Article

[65] PubMed/NCBI

[66] Google Scholar

[ref18] 18. Paterson C, Clevers H, Bozic I. Mathematical model of colorectal cancer initiation. Proc Natl Acad Sci U S A. 2020;117(34): 20681–20688. pmid:32788368
View Article
PubMed/NCBI
Google Scholar

[68] View Article

[69] PubMed/NCBI

[70] Google Scholar

[ref19] 19. Haupt S, Zeilmann A, Ahadova A, Blker H, Heuveline V. Mathematical modeling of multiple pathways in colorectal carcinogenesis using dynamical systems with kronecker structure. PLoS Comput Biol. 2021;17(5): e1008970. pmid:34003820
View Article
PubMed/NCBI
Google Scholar

[72] View Article

[73] PubMed/NCBI

[74] Google Scholar

[ref20] 20. Zhang R, Ukogu OA, Bozic I. Waiting times in a branching process model of colorectal cancer initiation. Theor Popul Biol. 2023;151: 44–63. pmid:37100121
View Article
PubMed/NCBI
Google Scholar

[76] View Article

[77] PubMed/NCBI

[78] Google Scholar

[ref21] 21. Lengauer C, Kinzler KW, Vogelstein B. Genetic instabilities in human cancers. Nature. 1998;396(6712): 643–649. pmid:9872311
View Article
PubMed/NCBI
Google Scholar

[80] View Article

[81] PubMed/NCBI

[82] Google Scholar

[ref22] 22. Wu Q, Smith-Miles K, Tian T. Approximate Bayesian computation schemes for parameter inference of discrete stochastic models using simulated likelihood density. BMC Bioinform. 2014; 15(Suppl 12): S3. pmid:25473744
View Article
PubMed/NCBI
Google Scholar

[84] View Article

[85] PubMed/NCBI

[86] Google Scholar

[ref23] 23. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA Jr, Kinzler KW. Cancer genome landscapes. Science. 2013;339(6127): 1546–1558. pmid:23539594
View Article
PubMed/NCBI
Google Scholar

[88] View Article

[89] PubMed/NCBI

[90] Google Scholar

[ref24] 24. Durrett R. Branching Process Models of Cancer. In: Branching Process Models of Cancer. Mathematical Biosciences Institute Lecture Series, vol 1.1. Springer, Cham; 2015.

[ref25] 25. Harris TE. The theory of branching processes. 1st ed. Springer-Verlag Berlin Heidelberg; 1963.

[ref26] 26. Portier CJ, Sherman CKA. Calculating tumor incidence rates in stochastic models of carcinogenesis. Math Biosci. 1996;135(2): 129–146. pmid:8768218
View Article
PubMed/NCBI
Google Scholar

[94] View Article

[95] PubMed/NCBI

[96] Google Scholar

[ref27] 27. Tomasetti C, Vogelstein B. Cancer etiology. Variation in cancer risk among tissues can be explained by the number of stem cell divisions. Science. 2015;347(6217): 78–81. pmid:25554788
View Article
PubMed/NCBI
Google Scholar

[98] View Article

[99] PubMed/NCBI

[100] Google Scholar

[ref28] 28. Komarova NL, Wodarz D. Drug resistance in cancer: principles of emergence and prevention. Proc Natl Acad Sci U S A. 2005;102(27): 9714–9719. pmid:15980154
View Article
PubMed/NCBI
Google Scholar

[102] View Article

[103] PubMed/NCBI

[104] Google Scholar

[ref29] 29. Shamieh SE, Saleh F, Assaad S, Farhat F. Next-generation sequencing reveals mutations in RB1, CDK4 and TP53 that may promote chemo-resistance to palbociclib in ovarian cancer. Drug Metabol Therapy. 2019;34(2): 20180027. pmid:31145688
View Article
PubMed/NCBI
Google Scholar

[106] View Article

[107] PubMed/NCBI

[108] Google Scholar

[ref30] 30. Lièvre A, Bachet JB, Corre DL, Boige V, Landi B, Emile JF, et al. KRAS mutation status is predictive of response to cetuximab therapy in colorectal cancer. Cancer Res. 2012;66(8): 3992–3995.
View Article
Google Scholar

[110] View Article

[111] Google Scholar

[ref31] 31. Pao W, Wang TY, Riely GJ, Miller VA, Pan Q, Ladanyi M, et al. KRAS mutations and primary resistance of lung adenocarcinomas to gefitinib or erlotinib. PLoS Medicine. 2005;2(1): e17. pmid:15696205
View Article
PubMed/NCBI
Google Scholar

[113] View Article

[114] PubMed/NCBI

[115] Google Scholar

[ref32] 32. Luebeck EG, Curtius K. Impact of tumor progression on cancer incidence curves. Cancer Res. 2013;73(3): 1086–1096. pmid:23054397
View Article
PubMed/NCBI
Google Scholar

[117] View Article

[118] PubMed/NCBI

[119] Google Scholar

[ref33] 33. Lamlum H, Papadopoulou A, Ilyas M, Rowan A, Gillet C, Hanby A, et al. APC mutations are sufficient for the growth of early colorectal adenomas. Proc Natl Acad Sci U S A. 2000;97(5): 2225–2228. pmid:10681434
View Article
PubMed/NCBI
Google Scholar

[121] View Article

[122] PubMed/NCBI

[123] Google Scholar

[ref34] 34. Snippert HJ, Schepers AG, van Es JH, Simons BD, Clevers H. Biased competition between Lgr5 intestinal stem cells driven by oncogenic mutation induces clonal expansion. EMBO Rep. 2014;15(1): 62–69. pmid:24355609
View Article
PubMed/NCBI
Google Scholar

[125] View Article

[126] PubMed/NCBI

[127] Google Scholar

[ref35] 35. Baker AM, Cereser B, Melton S, Fletcher AG, Rodriguez-Justo M, Tadrous PJ, et al. Quantification of crypt and stem cell evolution in the normal and neoplastic human colon. Cell Rep. 2014;8(4): 940–947. pmid:25127143
View Article
PubMed/NCBI
Google Scholar

[129] View Article

[130] PubMed/NCBI

[131] Google Scholar

[ref36] 36. Nicholson AM, Olpe C, Hoyle A, Thorsen AS, Rus T, Colombé M, et al. Fixation and spread of somatic mutations in adult human colonic epithelium. Cell Stem Cell. 2018;22(6): 909–918. pmid:29779891
View Article
PubMed/NCBI
Google Scholar

[133] View Article

[134] PubMed/NCBI

[135] Google Scholar

[ref37] 37. Vermeulen L, Morrissey E, van der Heijden D, Nicholson AM, Sottoriva A, Buczacki S, et al. Defining stem cell dynamics in models of intestinal tumor initiation. Science. 2013;342(6161): 995–998. pmid:24264992
View Article
PubMed/NCBI
Google Scholar

[137] View Article

[138] PubMed/NCBI

[139] Google Scholar

[ref38] 38. Tomasetti C, Bozic I. The (not so) immortal strand hypothesis. Stem Cell Res. (Amst.) 2015;14(2): 238–241. pmid:25700960
View Article
PubMed/NCBI
Google Scholar

[141] View Article

[142] PubMed/NCBI

[143] Google Scholar

[ref39] 39. Blokzijl F, Ligt JD, Jager M, Sasselli V, Roerink S, Sasaki N, et al. Tissue-specific mutation accumulation in human adult stem cells during life. Nature. 2016;538(7624): 260–264. pmid:27698416
View Article
PubMed/NCBI
Google Scholar

[145] View Article

[146] PubMed/NCBI

[147] Google Scholar

[ref40] 40. Fukasawa K, Choi T, Kuriyama R, Rulong S, Vande Woude GF. Abnormal centrosome amplification in the absence of p53. Science. 1996;271(5256): 1744–1747. pmid:8596939
View Article
PubMed/NCBI
Google Scholar

[149] View Article

[150] PubMed/NCBI

[151] Google Scholar

[ref41] 41. Fodde R, Kuipers J, Rosenberg C, Smits R, Kielman M, Gaspar C, et al. Mutations in the APC tumour suppressor gene cause chromosomal instability. Nature Cell Biol. 2001;3(4): 433–438. pmid:11283620
View Article
PubMed/NCBI
Google Scholar

[153] View Article

[154] PubMed/NCBI

[155] Google Scholar

[ref42] 42. Timmermann B, Kerick M, Roehr C, Fischer A, Isau M, Boerno ST, et al. Somatic mutation profiles of MSI and MSS colorectal cancer identified by whole exome next generation sequencing and bioinformatics analysis. PLoS One. 2010;5(12): e15661. pmid:21203531
View Article
PubMed/NCBI
Google Scholar

[157] View Article

[158] PubMed/NCBI

[159] Google Scholar

[ref43] 43. Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature. 2012;487(7407): 330–337.
View Article
Google Scholar

[161] View Article

[162] Google Scholar

[ref44] 44. Powell SM, Zilz N, Beazer-Barclay Y, Bryan TM, Hamilton SR, Thibodeau SN, et al. APC mutations occur early during colorectal tumorigenesis. Nature. 1992;359(6392): 235–237. pmid:1528264
View Article
PubMed/NCBI
Google Scholar

[164] View Article

[165] PubMed/NCBI

[166] Google Scholar

[ref45] 45. Suraweera N, Duval A, Reperant M, Vaury C, Furlan D, Leroy K, et al. Evaluation of tumor microsatellite instability using five quasimonomorphic repeats and pentaplex PCR. Gastroenterology. 2002;123(6): 1804–1811.
View Article
Google Scholar

[168] View Article

[169] Google Scholar

[ref46] 46. Zhang X, Fang Y, Zhao YD, Zheng WM. Mathematical modeling the pathway of human breast cancer. Math Biosci. 2014;253: 25–29. pmid:24680645
View Article
PubMed/NCBI
Google Scholar

[171] View Article

[172] PubMed/NCBI

[173] Google Scholar

[ref47] 47. Paterson C, Bozic I, Smith MJ, Hoad X, Evans DGR. A mechanistic mathematical model of initiation and malignant transformation in sporadic vestibular schwannoma. Br J Cancer. 2022;127: 1843–1857. pmid:36097176
View Article
PubMed/NCBI
Google Scholar

[175] View Article

[176] PubMed/NCBI

[177] Google Scholar

[ref48] 48. Golightly A, Wilkinson DJ. Bayesian parameter inference for stochastic biochemical network models using particle Markov chain Monte Carlo. Interface Focus. 2011;1(6): 807–820. pmid:23226583
View Article
PubMed/NCBI
Google Scholar

[179] View Article

[180] PubMed/NCBI

[181] Google Scholar

[ref49] 49. Bollen Y, Stelloo E, van Leenen P, van den Bos M, Ponsioen B, Lu BX, et al. Reconstructing single-cell karyotype alterations in colorectal cancer identifies punctuated and gradual diversification patterns. Nat Genet. 2021;53(8): 1187–1195. pmid:34211178
View Article
PubMed/NCBI
Google Scholar

[183] View Article

[184] PubMed/NCBI

[185] Google Scholar

[ref50] 50. Caravagna G, Graudenzi A, Ramazzotti D, Sanz-Pamplona R, Sano LD, Mauri G, et al. Algorithmic methods to infer the evolutionary trajectories in cancer progression. Proc Natl Acad Sci U S A. 2016;113(28): E4025–E4034 pmid:27357673
View Article
PubMed/NCBI
Google Scholar

[187] View Article

[188] PubMed/NCBI

[189] Google Scholar

[ref51] 51. Fleck JL, Pavel AB, Cassandras CG. Integrating mutation and gene expression cross-sectional data to infer cancer progression. BMC Syst Biol. 2016;10(1): 1–12. pmid:26810975
View Article
PubMed/NCBI
Google Scholar

[191] View Article

[192] PubMed/NCBI

[193] Google Scholar

[ref52] 52. Diaz-Colunga J, Diaz-Uriarte R. Conditional prediction of consecutive tumor evolution using cancer progression models: What genotype comes next? PLoS Comput Biol. 2021;17(12): e1009055–e1009055. pmid:34932572
View Article
PubMed/NCBI
Google Scholar

[195] View Article

[196] PubMed/NCBI

[197] Google Scholar

Figures

Abstract

Author summary

Introduction

Materials and methods

Results

Comparison of two solutions

The order of gene alterations

The number of cells with different genetic alterations

Discussion

Supporting information

S1 Text. Mathematical derivation and supplementary figures.

Acknowledgments

References