Aside from primary prevention, early detection remains the most effective way to decrease mortality associated with the majority of solid cancers. Previous cancer screening models are largely based on classification of at-risk populations into three conceptually defined groups (normal, cancer without symptoms, and cancer with symptoms). Unfortunately, this approach has achieved limited successes in reducing cancer mortality. With advances in molecular biology and genomic technologies, many candidate somatic genetic and epigenetic “biomarkers” have been identified as potential predictors of cancer risk. However, none have yet been validated as robust predictors of progression to cancer or shown to reduce cancer mortality. In this Perspective, we first define the necessary and sufficient conditions for precise prediction of future cancer development and early cancer detection within a simple physical model framework. We then evaluate cancer risk prediction and early detection from a dynamic clonal evolution point of view, examining the implications of dynamic clonal evolution of biomarkers and the application of clonal evolution for cancer risk management in clinical practice. Finally, we propose a framework to guide future collaborative research between mathematical modelers and biomarker researchers to design studies to investigate and model dynamic clonal evolution. This approach will allow optimization of available resources for cancer control and intervention timing based on molecular biomarkers in predicting cancer among various risk subsets that dynamically evolve over time.
Citation: Li X, Blount PL, Vaughan TL, Reid BJ (2011) Application of Biomarkers in Cancer Risk Management: Evaluation from Stochastic Clonal Evolutionary and Dynamic System Optimization Points of View. PLoS Comput Biol 7(2): e1001087. https://doi.org/10.1371/journal.pcbi.1001087
Editor: Claus O. Wilke, University of Texas at Austin, United States of America
Published: February 24, 2011
Copyright: © 2011 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.
Funding: Supported by NIH P01 CA91955 and RC1 CA146973 grants. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.
Competing interests: The authors have declared that no competing interests exist.
Detection of cancer at an early stage could significantly reduce cancer mortality and the overall burden of cancer –. The most common cancer risk model is based on the classification of the population into three groups: (1) normal without cancer, (2) asymptomatic cancer (detectable and potentially curable), and (3) symptomatic cancer, , . This model has provided a foundation for clinical approaches to cancer screening and early detection that have largely been based on tissue morphological features observed microscopically or via imaging. Recent advances in cancer biology and molecular technology now provide new opportunities to further refine cancer risk models for use in the clinic.
Inherited susceptibility and environmental exposures, including infectious agents, can modulate cancer risk in an individual over time, but dynamic interactions between these factors and evolving somatic genetic abnormalities that lead to cancer are poorly understood. With progress in molecular biology and genetics, it is widely believed that a panel of biomarkers assessing DNA, RNA, proteins, and/or metabolic processes can eliminate the shortcomings of morphologic diagnosis for early detection and cancer risk prediction. In fact, advances in technology have identified many molecular abnormalities that develop during neoplastic progression, some of which are highly associated with cancer –. Currently, there is a quest to find the perfect cancer biomarker(s) that could be used to separate “cancer” from “non-cancer”. Yet, thus far, no molecular biomarkers that significantly reduce cancer mortality with satisfactory sensitivity and specificity have become widely used in the clinic for early diagnosis or cancer risk prediction in the general population, although some genetic tests have been adopted for individuals with inherited susceptibilities to cancer . The limited success in identifying robust biomarkers has been attributed to inadequate study designs or complexity of biospecimens , , biased biospecimens , and technologic ,  and computational limitations –. Although all of these reasons contribute to the limited success of cancer biomarker development to some degree, a fundamental challenge to be considered for biomarker development is the dynamic, stochastic nature of clonal evolution.
In this paper, we first define cancer “risk prediction” and “early cancer detection” using two very simple physical systems. We then introduce the dynamic, clonal evolutionary concept as it relates to biomarkers in cancer development and review the challenges associated with the use of molecular markers in a clinical model. Finally, we present a general theoretical framework for using imperfect biomarkers for cancer risk stratification in the clinic, using a dynamic systems approach. The goal of this Perspective is to inspire an integrated approach to theoretical modeling and biomarker research in cancer risk management by incorporating a clonal dynamic point of view.
Event Detection and Prediction in Deterministic and Stochastic Systems
In order to better illustrate the problems associated with event prediction and early detection, we will introduce two simple physical systems as shown in Figures 1 and 2. For the purposes of this illustration, we are interested in determining (1) when the light bulb will turn on (event prediction) or (2) whether the light bulb is already on (event detection).
In this system, an on-off switch, S1, is connected to a power supply and timer A. Timer A and B or U are each connected to a switch (S2, S3, or S3′) that will automatically connect the wire when the timers count down to 0. Switch S3 or S3′ is connected to cumulative timer C, which counts cumulative time since S3 or S3′ is switched on and which is connected to a light bulb that is situated in a metal box (whether the bulb is on or off is not observable directly). Timers A, B (and U), and C are set to their initial position at 40, 60 (and 80), and 0 seconds, respectively. Timer A starts to count down after S1 is set to the “on” position by turning power on. Timer A will count down 40 seconds, which will trigger S2 to switch to on position, which then triggers timer B and U to start counting down (pathway U is used to represent a possible alternative pathway). After timer B runs 60 seconds or U runs 80 seconds, it will set S3 (or S3′) to be connected, respectively, which will trigger timer C to count down and at the same time the light bulb will turn on. If one path (either B/S3 or U/S3′) is blocked, then the other one could still function to turn the light bulb on. Event prediction can be made with 100% certainty by observing timers A, U, or B before the light bulb turns on and event detection can be determined with equal accuracy by observing cumulative timer C after the light bulb is turned on.
Figure 2 is similar to Figure 1 except that switches S2 and S3 (or S3′) have probabilities of P'1 and P'2 (P'3) being switched to dead ends that will never lead to a pathway that causes the light bulb to turn on (pathway U is used to represent a possible alternative pathway). The timers (A, B, and U) are also of random nature. In this system, due to random nature of S2 and S3 (or S3′) and timers, one will never be able to determine with precise accuracy when the light bulb will turn on, or whether it will turn on at all.
Event Prediction and Event Detection in a Deterministic System
We first investigate the question of event prediction in the deterministic system (for any given time point or initial state, the system will always produce the same outcome) as illustrated in Figure 1. In this system, all of the switches (S1, S2, S3, and S3′) have been precisely identified and characterized and the timers in the system (A, B, and U) are 100% accurate and clearly observable. In this deterministic model, the status of the light bulb can always be predicted perfectly by simply observing timers A, B, and U (see Figure 1 legend for details). Therefore, the necessary and sufficient conditions for 100% successful event prediction in this system are (1) all of the switches in the system are identified and characterized, and (2) the timers reflect the exact time (sufficient conditions) that has elapsed since any two switches have connected and are observable.
The question of whether the event has already occurred (light bulb is on) and the time elapsed after occurrence (event detection) can be answered in Figure 1 by simply observing cumulative timer C. Alternatively, it could be determined by observing the status of switch(es) S3, (S3′). Thus, the necessary and sufficient conditions for event detection can be determined by simply observing the status of the final switches or cumulative timer C.
Event Prediction and Event Detection in a Stochastic System
A stochastic system that may not always result in the same outcome for any given time point or initial state is shown in Figure 2. In this system, one cannot make a prediction with 100% accuracy as to when the light bulb will turn on due to the stochastic nature of switches S2 and S3 (or S3′) and timers A, B, and U. In order to make a reasonably good prediction as to when the light bulb will turn on, one will need to periodically recheck the status of the switches.
The “switches” in Figures 1 and 2 may be viewed as similar to the abnormalities (e.g., DNA mutations/LOH/chromosome copy gain/loss) that develop during neoplastic progression, and the timers as the time that has elapsed between the genomic alterations. The event of turning on the light bulb may be intuitively analogous to the development of cancer. However, neither of these models addresses the complexity of neoplastic progression.
Biological Complexity of Molecular Markers for Early Detection of Cancer and Cancer Risk Prediction
Environmental factors, inherited genetic susceptibility, and dynamic gene–gene and gene–environmental interactions can generate new somatic genetic or epigenetic alterations in human cells. Some altered cells continue to divide with a low rate of genetic or epigenetic alterations, largely maintaining genomic integrity, whereas others continue to develop new alterations, generating new variants on which selection can act. Several investigators have used genetic and genomic data to develop models of neoplastic progression. Nowell proposed that a single cell gains a selective genetic advantage over neighboring normal cells, rendering it “neoplastic” . Genomic instability and stepwise selection of genetic variants lead to clonal expansions of viable lineages of daughter cells in somatic tissue and underlies solid tumor progression while most variant cells or clones do not evolve to cancer. Later, Fearon and Vogelstein  first proposed a sequence of key genetic events during the evolution of a normal colon cell to colon cancer. Barrett et al.  reported the evolution of neoplastic lineages in Barrett's esophagus using spatial and temporal data in individual patients, including ordering of early key genetic events, as well as tracing genetic lineages of viable clones that produced a cancer and clones that were not selected for progression to cancer. Based on molecular data and population cancer incidence data, a number of stochastic and mathematical models have been proposed to evaluate mechanisms of initiation and development of cancer –. In addition, a stochastic system dynamic modeling approach was utilized to study the mechanisms of cancer genesis and progression by using combined data from various molecular biology studies . The stochastic properties of selected and nonselected chromosomal abnormalities in hyper-dynamic evolution was further illustrated by a cell population heterogeneity study in which the highest level of non-clonal chromosome aberrations was closely coupled with the strongest tumorigenicity .
Many studies have shown that advanced epithelial malignancies have typically accumulated large numbers of genomic abnormalities not found in normal cells, including whole or segmental chromosome copy number amplifications, deletions, loss of heterozygosity, translocations, and point mutations –. Jones et al.  used spatial data from individual patients with colorectal cancers and reported that times between benign, invasive, and metastatic colon tumors can be estimated by analysis of the mutations they have in common and knowledge of the time it takes for cell division. Although these biological and genetic models provide significant insights for understanding evolution of a normal cell to cancer, all of them have stochastic characteristics (similar to Figure 2), and none of them meet the conditions required for perfect cancer risk prediction. This is because the exact steps and precise time elapsed between each step cannot be predicted with 100% accuracy in contrast to Figure 1. As a consequence, in clinical practice we will likely have imperfect biomarkers to identify high risk persons for targeted prevention strategies (cancer risk prediction) and those with early stage curable cancer for treatment (combination of cancer risk prediction and early detection). We next propose a schematic model that could be used for modeling cancer risk management with consideration of stochastic characteristics in the evolution of cancer (Figure 3) and more accurate risk stratification using current molecular measurements, as illustrated in Figure 4.
(A), (B), and (C) represent evolutionary stages in dynamic clonal progression to cancer. (A) Early stages of clonal evolution have fewer selected genomic alterations, and most individuals do not progress to cancer. (B) A minority of individuals will evolve additional genomic alterations, but the majority of these will not progress to cancer. (C) A small subset of patients will accelerate development of genomic alterations leading to selection of increasing abnormal clones and progression to cancer. These events are stochastic and there are no biomarkers that perfectly distinguish (A), (B), and (C). In this evolutionary process, most clones may evolve in directions that do not lead to cancer (dark gray circles), whereas some others retain great potential for future progression to cancer or development of resistance to interventions to prevent or treat cancer (white circles), depending on selective pressures. Only a minority of the evolving clones will eventually acquire the capacity to become cancerous (red circles), and progression to cancer can occur by multiple possible pathways, also illustrated in the figure. The initial cancer cells may continue to divide locally and produce future metastases (purple circles in red block). The gray bar surrounded by dashed lines at the bottom of the figure illustrates the use of biomarkers in a clonal evolutionary system. Biomarkers with increasing specificity and sensitivity would shrink the gray block from either side toward the small center red bar, at the conceptual transition between non-cancer and cancer. A biomarker with perfect sensitivity and specificity would exactly correspond to the position of the red bar with perfect separation of cancer and non-cancer. Thus far, no biomarkers satisfy the necessary and sufficient conditions for precise cancer risk prediction or early cancer detection (as in Figure 1).
Clinical patients or the general population could be stratified objectively using molecular biomarkers. A set of robust and validated biomarkers is expected to stratify the majority of patients into either high or low risk groups, and assign fewer patients to the intermediate risk group. The cancer risks of each group could be managed with consideration of individual risk, risk-benefit assessment, quality of life available resources, and dynamic progression characteristics of each group to achieve overall optimized results.
Cancer Risk Prediction, Early Detection and Management: A Dynamic Systems Point of View
Modeling the System
Previous models , , ,  developed for cancer screening and early detection have been based on the definition of three conceptual stages: normal, cancer without symptoms, and cancer with symptoms. The analysis of screening for early detection of cancer was carried out using assumptions about the statistical distributions of transition from non-cancer to cancer and sojourn time. The approach of using three conceptual stages for classification of populations for modeling screening is reasonable. However, these models can not be easily related to, or characterized by, the complexity of clonal evolution as reported by recent genomic studies using high throughput sequencing or high density arrays.
One obvious approach is to use molecular or genomic biomarkers to stratify populations into various risk groups where each group has its own cancer risk distribution rather than treating overall cancer risk as a single distribution pattern in a population. However, molecular biomarkers are likely to be imperfect and have stochastic characteristics (Figure 3), and thus far, none of them meet the conditions required for perfect cancer risk prediction. In practice, one will be using imperfect biomarkers for both risk prediction and early detection. As shown in Figure 4, biomarkers could be used to stratify populations into different risk groups –. Each risk group could then be individually managed, with the goal of reducing the overall cancer risk over time, given a set of fixed resources or time constraints.
Our framework for cancer risk management considers the situation in which members in lower risk groups could progress to higher risk groups over time or, alternatively, members in higher risk groups could regress to a lower risk status. Such dynamic risk status will need to be periodically re-evaluated over time using biomarkers. Patients with cancer will be treated immediately, while a high risk group could be provided with an intervention to prevent cancer, such as chemoprevention, or monitored more closely for early detection. The lower risk groups could be subjected to much less intensive monitoring. Therefore, early detection and cancer risk management would comprehensively consider a dynamic system over a period of time that includes (1) the absolute risk for each risk group (cancer incidence rates), which is characterized by molecular biomarkers and estimated through population studies, (2) a pre-specified level of late stage cancer reduction, (3) resources available for cancer risk reduction or management, (4) the quality of an individual's life (effects of false positive or false negative diagnosis on quality of life), and (5) competing mortality due to other causes as shown in Figure 4.
There are many potential mathematical approaches to quantify or model such a dynamic system (Figure 4) for cancer risk management. In this Perspective, we suggest one possible method using dynamic system optimization to deal with clonal evolution for cancer risk management using biomarkers (Text S1).
Discussion and Concluding Remarks
The development of cancer is a complex process characterized by stochastic accumulation of mutations and dynamic evolution of clones , , –. To date, most biomarkers are not directly related to evolutionary dynamics, but instead focus on specific pathways. However, extrapolating information from pathways to evolutionary dynamics remains a challenge . Recently, genome-based cell population heterogeneity rather than commonly shared pathways have been linked to tumorigenicity . Whole genome instability measurements are also closely linked to stages of cancer progression . This body of evidence indicates that useful biomarkers can be developed directly using longitudinal measurements of genomic instability.
Development of biomarkers for cancer risk management should consider these stochastic and dynamic properties over time in neoplastic evolution. Using biomarkers for cancer risk management involves multi-level systems from cells to individuals to populations. Better ways of modeling multi-level systems and comprehending innate uncertainties in these systems are areas in which great benefits could be achieved . We propose a dynamic system optimization approach to deal with the practical limitations of earlier three-stage models (Figure 4 and Text S1).
The goal of this Perspective is to link the dynamic, stochastic elements of clonal evolution in neoplastic tissues in patients followed over time. This would be best accomplished by collaborations between mathematical modelers and laboratory researchers in longitudinal experimental design, modeling, and parameter estimation oriented to practical application for cancer risk management.
There are fundamental differences between our concept and the commonly used three-stage (normal, cancer without symptoms, and cancer with symptoms) model for cancer screening: (1) we evaluate cancer risk prediction and early detection from a dynamic clonal evolutionary point of view and its implication for cancer risk management in clinical practice; (2) based on stochastic clonal evolution, we propose a framework to guide future biomarker research to more accurately stratify patients into various risk groups, where each risk group has a different cancer risk distribution, thereby permitting an adaptive cancer risk strategy; and (3) we propose that mathematical models be developed for cancer risk management that can be expanded for modeling specific cancer or management aspects, allowing optimization of available resources and intervention timing based on particular biomarker sensitivity and specificity in predicting cancer among various risk groups that dynamically evolve over time (Text S1). In addition, points (2) and (3) above take into account clonal dynamics in cancer development, including both progression and regression depending on dynamic selective conditions. Models using biomarkers for cancer risk management should allow optimization of various specific and objective functions dynamically, which is different from the traditional medical decision tree approach. Note that we addressed intrinsic stochastic properties of clonal dynamics in cancer development, as distinct from issues such as noise due to measurement error in biomarker studies.
In summary, effective prevention and early diagnosis strategies are critical to reducing the cancer burden. We analyzed the biological basis for using biomarkers for cancer risk prediction and early detection. This analysis shows that if and only if biological pathways for cancer development are fully determined and quantified as in Figure 1, can perfect accuracy for cancer risk prediction and early detection be achieved. With the advances of molecular technology and knowledge, biomarkers could reach a high level of accuracy in cancer risk prediction and early detection and could be used to guide clinical monitoring and interventions but likely will never be perfectly accurate. We propose that evaluation of biomarker effectiveness for cancer risk prediction and early detection be conducted with consideration of cancer evolutionary dynamics and dynamic optimization modeling for risk management.
A dynamic optimization approach for cancer risk management and early detection. In S1, we present a framework of mathematical modeling for cancer risk management using biomarkers with consideration of stochastic clonal evolution in neoplastic progression.
(0.03 MB DOC)
- 1. Wingo PA, Ries LA, Rosenberg HM, Miller DS, Edwards BK (1998) Cancer incidence and mortality, 1973–1995: a report card for the U.S. Cancer 82: 1197–1207.PA WingoLA RiesHM RosenbergDS MillerBK Edwards1998Cancer incidence and mortality, 1973–1995: a report card for the U.S.Cancer8211971207
- 2. Etzioni R, Urban N, Ramsey S, McIntosh M, Schwartz S, et al. (2003) The Case for Early Detection. Nature Reviews Cancer Volume 3: 1–10.R. EtzioniN. UrbanS. RamseyM. McIntoshS. Schwartz2003The Case for Early Detection.Nature Reviews Cancer Volume3110
- 3. Jemal A, Clegg LX, Ward E, Ries LA, Wu X, et al. (2004) Annual report to the nation on the status of cancer, 1975–2001, with a special feature regarding survival. Cancer 101: 3–27.A. JemalLX CleggE. WardLA RiesX. Wu2004Annual report to the nation on the status of cancer, 1975–2001, with a special feature regarding survival.Cancer101327
- 4. O'Connell JB, Maggard MA, Ko CY (2004) Colon cancer survival rates with the new American Joint Committee on Cancer sixth edition staging. J Natl Cancer Inst 96: 1420–1425.JB O'ConnellMA MaggardCY Ko2004Colon cancer survival rates with the new American Joint Committee on Cancer sixth edition staging.J Natl Cancer Inst9614201425
- 5. Zelen M (1993) Optimal scheduling of examinations for the early detection of disease. Biometrika 80: 279–293.M. Zelen1993Optimal scheduling of examinations for the early detection of disease.Biometrika80279293
- 6. Prorok PC (1976) The theory of periodic screening I. Lead time and proportion detected. Advances in applied probability 8: 127–143.PC Prorok1976The theory of periodic screening I. Lead time and proportion detected.Advances in applied probability8127143
- 7. Croce CM (2008) Oncogenes and cancer. N Engl J Med 358: 502–511.CM Croce2008Oncogenes and cancer.N Engl J Med358502511
- 8. Esteller M (2008) Epigenetics in cancer. N Engl J Med 358: 1148–1159.M. Esteller2008Epigenetics in cancer.N Engl J Med35811481159
- 9. Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell 100: 57–70.D. HanahanRA Weinberg2000The hallmarks of cancer.Cell1005770
- 10. Jones S, Chen WD, Parmigiani G, Diehl F, Beerenwinkel N, et al. (2008) Comparative lesion sequencing provides insights into tumor evolution. Proc Natl Acad Sci U S A 105: 4283–4288.S. JonesWD ChenG. ParmigianiF. DiehlN. Beerenwinkel2008Comparative lesion sequencing provides insights into tumor evolution.Proc Natl Acad Sci U S A10542834288
- 11. Robson ME, Storm CD, Weitzel J, Wollins DS, Offit K (2010) American Society of Clinical Oncology policy statement update: genetic and genomic testing for cancer susceptibility. J Clin Oncol 28: 893–901.ME RobsonCD StormJ. WeitzelDS WollinsK. Offit2010American Society of Clinical Oncology policy statement update: genetic and genomic testing for cancer susceptibility.J Clin Oncol28893901
- 12. Diamandis EP (2010) Cancer biomarkers: can we turn recent failures into success? J Natl Cancer Inst 102: 1462–1467.EP Diamandis2010Cancer biomarkers: can we turn recent failures into success?J Natl Cancer Inst10214621467
- 13. Hanash SM, Pitteri SJ, Faca VM (2008) Mining the plasma proteome for cancer biomarkers. Nature 452: 571–579.SM HanashSJ PitteriVM Faca2008Mining the plasma proteome for cancer biomarkers.Nature452571579
- 14. Ransohoff DF (2005) Bias as a threat to the validity of cancer molecular-marker research. Nat Rev Cancer 5: 142–149.DF Ransohoff2005Bias as a threat to the validity of cancer molecular-marker research.Nat Rev Cancer5142149
- 15. Gerges N, Rak J, Jabado N (2010) New technologies for the detection of circulating tumour cells. Br Med Bull 94: 49–64.N. GergesJ. RakN. Jabado2010New technologies for the detection of circulating tumour cells.Br Med Bull944964
- 16. Dettling M, Buhlmann P (2003) Boosting for tumor classification with gene expression data. Bioinformatics 19: 1061–1069.M. DettlingP. Buhlmann2003Boosting for tumor classification with gene expression data.Bioinformatics1910611069
- 17. Handl J, Knowles J, Kell DB (2005) Computational cluster validation in post-genomic data analysis. Bioinformatics 21: 3201–3212.J. HandlJ. KnowlesDB Kell2005Computational cluster validation in post-genomic data analysis.Bioinformatics2132013212
- 18. Natsoulis G, El Ghaoui L, Lanckriet GR, Tolley AM, Leroy F, et al. (2005) Classification of a large microarray data set: algorithm comparison and analysis of drug signatures. Genome Res 15: 724–736.G. NatsoulisL. El GhaouiGR LanckrietAM TolleyF. Leroy2005Classification of a large microarray data set: algorithm comparison and analysis of drug signatures.Genome Res15724736
- 19. Nowell PC (1976) The clonal evolution of tumor cell populations. Science 194: 23–28.PC Nowell1976The clonal evolution of tumor cell populations.Science1942328
- 20. Fearon ER, Vogelstein B (1990) A genetic model for colorectal tumorigenesis. Cell 61: 759–767.ER FearonB. Vogelstein1990A genetic model for colorectal tumorigenesis.Cell61759767
- 21. Barrett MT, Sanchez CA, Prevo LJ, Wong DJ, Galipeau PC, et al. (1999) Evolution of neoplastic cell lineages in Barrett oesophagus. Nat Gen 22: 106–109.MT BarrettCA SanchezLJ PrevoDJ WongPC Galipeau1999Evolution of neoplastic cell lineages in Barrett oesophagus.Nat Gen22106109
- 22. Frank S (2007) Dynamics of Cancer: Incidence, Inheritance, and Evolution. Princeton: Princeton University Press. 378 p.S. Frank2007Dynamics of Cancer: Incidence, Inheritance, and Evolution.PrincetonPrinceton University Press378
- 23. Luebeck EG, Moolgavkar SH (2002) Multistage carcinogenesis and the incidence of colorectal cancer. Proc Natl Acad Sci U S A 99: 15095–15100.EG LuebeckSH Moolgavkar2002Multistage carcinogenesis and the incidence of colorectal cancer.Proc Natl Acad Sci U S A991509515100
- 24. Nowak MA, Komarova NL, Sengupta A, Jallepalli PV, Shih I-M, et al. (2002) The role of chromosomal instability in tumor initiation. Proceedings of the National Academy of Sciences of the United States of America 99: 16226–16231.MA NowakNL KomarovaA. SenguptaPV JallepalliI-M Shih2002The role of chromosomal instability in tumor initiation.Proceedings of the National Academy of Sciences of the United States of America991622616231
- 25. Tan WY (1991) Stochastic Models for Carcinogenesis. New York: Marcel Dekker, Inc. 264 p.WY Tan1991Stochastic Models for Carcinogenesis.New YorkMarcel Dekker, Inc264
- 26. Yakovlev AY, Tsodikov AD, Asselain B (1996) Stochastic models of tumor latency and their biostatistical applications. River Edge: World Scientific Publishing Co. 269 p.AY YakovlevAD TsodikovB. Asselain1996Stochastic models of tumor latency and their biostatistical applications.River EdgeWorld Scientific Publishing Co269
- 27. Ao P, Galas D, Hood L, Yin L, Zhu XM (2010) Towards predictive stochastic dynamical modeling of cancer genesis and progression. Interdiscip Sci 2: 140–144.P. AoD. GalasL. HoodL. YinXM Zhu2010Towards predictive stochastic dynamical modeling of cancer genesis and progression.Interdiscip Sci2140144
- 28. Ye CJ, Stevens JB, Liu G, Bremer SW, Jaiswal AS, et al. (2009) Genome based cell population heterogeneity promotes tumorigenicity: the evolutionary mechanism of cancer. J Cell Physiol 219: 288–300.CJ YeJB StevensG. LiuSW BremerAS Jaiswal2009Genome based cell population heterogeneity promotes tumorigenicity: the evolutionary mechanism of cancer.J Cell Physiol219288300
- 29. Li X, Galipeau PC, Sanchez CA, Blount PL, Maley CC, et al. (2008) Single nucleotide polymorphism-based genome-wide chromosome copy change, loss of heterozygosity, and aneuploidy in Barrett's esophagus neoplastic progression. Cancer Prev Res (Phila Pa) 1: 413–423.X. LiPC GalipeauCA SanchezPL BlountCC Maley2008Single nucleotide polymorphism-based genome-wide chromosome copy change, loss of heterozygosity, and aneuploidy in Barrett's esophagus neoplastic progression.Cancer Prev Res (Phila Pa)1413423
- 30. Wood LD, Parsons DW, Jones S, Lin J, Sjoblom T, et al. (2007) The genomic landscapes of human breast and colorectal cancers. Science 318: 1108–1113.LD WoodDW ParsonsS. JonesJ. LinT. Sjoblom2007The genomic landscapes of human breast and colorectal cancers.Science31811081113
- 31. Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, et al. (2007) Patterns of somatic mutation in human cancer genomes. Nature 446: 153–158.C. GreenmanP. StephensR. SmithGL DalglieshC. Hunter2007Patterns of somatic mutation in human cancer genomes.Nature446153158
- 32. Weir BA, Woo MS, Getz G, Perner S, Ding L, et al. (2007) Characterizing the cancer genome in lung adenocarcinoma. Nature 450: 893–898.BA WeirMS WooG. GetzS. PernerL. Ding2007Characterizing the cancer genome in lung adenocarcinoma.Nature450893898
- 33. Tanemura A, Terando AM, Sim MS, van Hoesel AQ, de Maat MF, et al. (2009) CpG island methylator phenotype predicts progression of malignant melanoma. Clin Cancer Res 15: 1801–1807.A. TanemuraAM TerandoMS SimAQ van HoeselMF de Maat2009CpG island methylator phenotype predicts progression of malignant melanoma.Clin Cancer Res1518011807
- 34. Beroukhim R, Mermel CH, Porter D, Wei G, Raychaudhuri S, et al. (2010) The landscape of somatic copy-number alteration across human cancers. Nature 463: 899–905.R. BeroukhimCH MermelD. PorterG. WeiS. Raychaudhuri2010The landscape of somatic copy-number alteration across human cancers.Nature463899905
- 35. Jones S, Zhang X, Parsons DW, Lin JC, Leary RJ, et al. (2008) Core signaling pathways in human pancreatic cancers revealed by global genomic analyses. Science 321: 1801–1806.S. JonesX. ZhangDW ParsonsJC LinRJ Leary2008Core signaling pathways in human pancreatic cancers revealed by global genomic analyses.Science32118011806
- 36. Flehinger BJ, Kimmel M (1987) The natural history of lung cancer in a periodically screened population. Biometrics 43: 127–144.BJ FlehingerM. Kimmel1987The natural history of lung cancer in a periodically screened population.Biometrics43127144
- 37. Lee SJ, Zelen M (1998) Scheduling periodic examinations for the early detection of disease: Applications to breast cancer. Journal of the American Statistical Association 93: 1271–1281.SJ LeeM. Zelen1998Scheduling periodic examinations for the early detection of disease: Applications to breast cancer.Journal of the American Statistical Association9312711281
- 38. Chen HY, Yu SL, Chen CH, Chang GC, Chen CY, et al. (2007) A five-gene signature and clinical outcome in non-small-cell lung cancer. N Engl J Med 356: 11–20.HY ChenSL YuCH ChenGC ChangCY Chen2007A five-gene signature and clinical outcome in non-small-cell lung cancer.N Engl J Med3561120
- 39. Galipeau PC, Li X, Blount PL, Maley CC, Sanchez CA, et al. (2007) NSAIDs modulate CDKN2A, TP53, and DNA content risk for future esophageal adenocarcinoma. PLoS Med 4: e67.PC GalipeauX. LiPL BlountCC MaleyCA Sanchez2007NSAIDs modulate CDKN2A, TP53, and DNA content risk for future esophageal adenocarcinoma.PLoS Med4e67
- 40. Lee JJ, Hong WK, Hittelman WN, Mao L, Lotan R, et al. (2000) Predicting cancer development in oral leukoplakia: ten years of translational research. Clin Cancer Res 6: 1702–1710.JJ LeeWK HongWN HittelmanL. MaoR. Lotan2000Predicting cancer development in oral leukoplakia: ten years of translational research.Clin Cancer Res617021710
- 41. Reid BJ, Li X, Galipeau PC, Vaughan TL (2010) Barrett's oesophagus and oesophageal adenocarcinoma: time for a new synthesis. Nat Rev Cancer 10: 87–101.BJ ReidX. LiPC GalipeauTL Vaughan2010Barrett's oesophagus and oesophageal adenocarcinoma: time for a new synthesis.Nat Rev Cancer1087101
- 42. Shedden K, Taylor JM, Enkemann SA, Tsao MS, Yeatman TJ, et al. (2008) Gene expression-based survival prediction in lung adenocarcinoma: a multi-site, blinded validation study. Nat Med 14: 822–827.K. SheddenJM TaylorSA EnkemannMS TsaoTJ Yeatman2008Gene expression-based survival prediction in lung adenocarcinoma: a multi-site, blinded validation study.Nat Med14822827
- 43. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AAM, et al. (2002) Gene expression profiling predicts clinical outcome of breast cancer. Nature 415: 530–535.LJ van 't VeerH. DaiMJ van de VijverYD HeAAM Hart2002Gene expression profiling predicts clinical outcome of breast cancer.Nature415530535
- 44. Heng HH, Stevens JB, Lawrenson L, Liu G, Ye KJ, et al. (2008) Patterns of genome dynamics and cancer evolution. Cell Oncol 30: 513–514.HH HengJB StevensL. LawrensonG. LiuKJ Ye2008Patterns of genome dynamics and cancer evolution.Cell Oncol30513514
- 45. Siegmunda KD, Marjorama P, Woob Y, Tavare S, Shibatab D (2008) Inferring clonal expansion and cancer stem cell dynamics from DNA methylation patterns in colorectal cancers. Proc Natl Acad Sci U S A 106: 4828–4833.KD SiegmundaP. MarjoramaY. WoobS. TavareD. Shibatab2008Inferring clonal expansion and cancer stem cell dynamics from DNA methylation patterns in colorectal cancers.Proc Natl Acad Sci U S A10648284833
- 46. Heng HH (2008) The conflict between complex systems and reductionism. JAMA 300: 1580–1581.HH Heng2008The conflict between complex systems and reductionism.JAMA30015801581
- 47. Sawyers CL (2008) The cancer biomarker problem. Nature 452: 548–552.CL Sawyers2008The cancer biomarker problem.Nature452548552
- 48. Cohen JE (2004) Mathematics is biology's next microscope, only better; biology is mathematics' next physics, only better. PLoS Biol 2: e439.JE Cohen2004Mathematics is biology's next microscope, only better; biology is mathematics' next physics, only better.PLoS Biol2e439