Cancer and the goals of integration
Introduction
In this paper, we will consider one central project in the history of cancer1 research, modeling carcinogenesis as a multi-stage process, as a case study for investigating the ideals of “unification” versus “integration” in the sciences. Multistage models of cancer represent cancer initiation and progression to neoplastic state as a multi-stage process, driven by the acquisition of a series of mutations. Sometimes this view is assimilated with an “evolutionary” perspective on cancer, since a cancer’s capacity to attract a blood supply, invade neighboring tissue, and metastasize, are all seen as the result of the acquisition of a series of mutations that increase the relative “fitness” of the cancer cells (Merlo et al., 2006, Nowell, 1976). Whether or no cancer progression is best viewed as an evolutionary process is a question requiring further exploration (Plutynski, in press); so, we focus here on the multistage theory.
Cancer incidence increases as a power of age; the multistage theory explains this phenomenon as due to the rate-limited accumulation of mutations to genes (as well as chromosomal and epigenetic changes) that play key roles in the regulation of the cell cycle. The theory also explains departures from average age of incidence curves. For instance, familial forms of cancer cause a shift the age of incidence curves, due to hereditary mutations that “accelerate” the onset of cancer (Knudson, 1971). The multistage theory appears to explain some patterns of cancer incidence quite well. Most notably, for colon cancer, not only have the specific series of mutations leading to a specific cancer type have been identified, but their mechanisms of action, and thus role in causing dysplastic growth, are well understood (Fearon & Vogelstein, 1990). The history of the multistage theory is a useful illustration of both the advantages and limits of mathematical modeling in arriving at general theories in biomedicine.
Carcinogenesis is a complex process, due to many causes acting both at the level of the cell and above (Bissell & Short, 2009). How, if at all, may a simple mathematical model capture all the various causes of cancer(s), acting at distinct temporal and spatial scales? It cannot, and it should not. Abstract models, such as the family of models of carcinogenesis, are intended to identify the central causal factors yielding some outcome, at one well-defined level of analysis. So, they deliberately exclude, for instance, causes exogenous to the system of interest. Such causes are treated as more or less a black box. At least initially, there was no representation in the multi-stage models of the role of the tumor microenvironment, immune system, diet, or smoking. Nonetheless, the models were a way of systematically representing carcinogenesis, consistent with a variety of independent evidence: patterns of cancer incidence by age, patterns of cancer incidence in childhood cancers, and toxicological data on the effects of chemical carcinogens on animals. More recently, such models have been integrated with data from molecular genetics on the role of specific genes in cell birth and death, and data on the rate and structure of cell renewal in different tissue types (Frank, 2007, Frank and Nowak, 2004). That is, what began as a way of modeling cancer at one level of analysis using simple mathematical models became a theoretical framework for integrating new data from different levels of analysis—both from the “bottom up” and “top down”.
Biologist Steven Frank (2007) calls the family of models that represent cancer as a multi-stage process the “dynamics” of cancer. This characterization suggests an analogy with Newtonian dynamics, the theory that unified terrestrial and celestial mechanics. Mathematical representations of cancer initiation and progression as a dynamic, multi-stage process are analogous to Newtonian mechanics in the following respects. They both treat complex phenomena using simple mathematical models, and both hypothesized that there were common causes, driving observed patterns. Moreover, both treat different phenomena—in the case of multistage theory, distinct cancers—as of a kind. Hereditary and somatic cancers, cancers found in different tissues or of different types are all, on this theory, subject to similar causes acting in similar ways. This unifying perspective had the virtue that it served as a guiding idea for a research program. Seeing distinct cancers as subject to similar causes was central to a research program that (in part) led to the discovery of a family of genes that play important roles in all cancers: TP53, RB, APC, HRAS. It also led to the realization that understanding genes and their activity in isolation from the tumor microenvironment was not sufficient to explain carcinogenesis. Thus, the history of multistage theory can serve as an interesting case study for the purported virtues of “integrative” or “unified” theories and explanations in the sciences.
As many philosophers writing on the explanatory power of unifying theories have noted (see Cartwright, 1980) unification often comes at a cost; unified theories or laws with wide scope trade generality and cohesiveness for simplification or omission of complex causal details. And, if anything is an instance of a complex causal processes, carcinogenesis is it. No single model could possibly incorporate all the factors affecting carcinogenesis; in part, because cancers are so different, but in part, also, because cancer is not simply a “genetic” disease. Mutations are a significant difference maker in cancer, but they are not the only one (Bissell & Hines, 2011). The multistage theory focuses causal factors that shift the age of incidence curve: core difference makers to the time of onset of cancer. This accumulation of mutations is taken to explain the fact that cancer incidence by and large increases as a power of age.
By focusing on mutations, the multistage theory trades simplicity and unifying power for explanatory detail. However, over time, the theory has come to incorporate evidence from a wider domain. In other words, the multistage model of cancer is a case study in integration as a process. While in some sense, it started as a “reductive” and “unifying” theory, reducing carcinogenesis to nothing more than the serial acquisition of genetic mutations, over time, it has incorporated more data, from a variety of methodological and theoretical perspectives. Though, even in the beginning, the theory required moving between levels of analysis; that is, it required the insight that epidemiological data on age of incidence might provide some clue as to the etiology of cancer.
While the sheer number and diversity of questions and subject matter in cancer research would seem to argue for greater specialization, recently, there has been a call from both granting agencies and major research universities for more “integrative” and “inter-” or “trans-disciplinary” research.2 As an example, in 2003, the US National Cancer Institute’s (NCI) Division of Cancer Biology “initiated a program to highlight (1) systems biology, (2) a systems approach to cancer biology, (3) interdisciplinary and collaborative research and (4) interdisciplinary training.” With an initial $14.9 million in funding, the Integrative Cancer Biology Program (ICBP) was created, and in 2004, nine interdisciplinary centers were founded, “incorporating a spectrum of new technologies such as genomics, proteomics, and molecular imaging, to generate computer and mathematical models that could predict the cancer process.” (NCI, 2012, http://icbp.nci.nih.gov/). “Integrative” research, in part as a result of funding initiatives such as the above, has been promoted in many other areas in biomedical science: many institutions have founded research programs, institutes, and centers of “integrative” research. A search in PubMed with the terms “cancer” and “integration” turns up over 9000 hits.
What exactly is being called for with these demands for integrative research? And why should we presume that integration would be a good thing? Perhaps because so many kinds of things may be “integrated” (data, methods, explanations; see O’Malley, 2013), there are many different meanings at play; and, all too often the term is used rhetorically as an advertisement of forward-thinking science with very little warrant. So, what does warrant the appellation? This paper will provide an overview of different philosophical accounts of both what integration might be (Section 2); a case study of the multistage theory (Section 3); and, finally an argument for the following: (a) the success of explanatory “integration” is always relative to some specific scientific problem (cf. Brigandt, 2010, Brigandt, 2013; also, Love, 2008a, Love, 2008b), and so one should be wary of generalizations about “the” goals of explanatory integration, and (b) theoretical frameworks may become successively more “integrative” over time, along a variety of dimensions (Section 4). While these points may be not entirely novel (see, e.g., Bechtel, 2010, Brigandt, 2013, Love, 2010), the below will hopefully provide some framework for future discussion of both the goals of integration in future philosophical work, particularly in the context of biomedical research.
Section snippets
What is integration? How is it distinct from unification?
A variety of philosophers of science have offered models of “integrative” research, and “integrative explanation” in overlapping problem domains, as well as “interfield” theories, as one alternative to the view that unification in the sciences is achieved viz. theory reduction in Nagel’s (1961) sense (cf. Maull, 1977, Darden and Maull, 1977, Darden, 2005, Bechtel, 1986, Bechtel, 1993, Bechtel, 2010, Mitchell, 2003, Mitchell and Dietrich, 2006, Craver, 2005, Craver, 2007, Brigandt, 2012,
Case study: multi-stage model of carcinogenesis
Starting in the 1940s, biologists began to develop a family of models that represent cancer initiation and progression mathematically as a product of a series of rate-limited steps. The first quantitative mathematical models to represent cancer as the product of multistage progression were developed by two biologists who found that mice acquired skin tumors after repeated application of benzopyrene (Charles & Luce-Clausen, 1942). Charles and Luce-Clawson hypothesized that cancer was the product
Conclusions
It is by now a truism that at least one of the aims of science, if not the central aim, is to understand the causal structure of the world (Salmon, 1989). In medicine, the aim is also (if not primarily) to change the world—to prevent and cure disease. To change the world, we must know how to intervene on the world. One may intervene effectively with less than complete causal understanding. Many medical interventions were known to be effective without any clear (and sometimes false)
Acknowledgements
Many thanks to Ingo Brigandt, Alan Love, Alex Broadbent, Jim Tabery, Danielle Endres, Christopher Hunter Lean, Matt Haber, Rob Gehl, Marta Bertolaso, the participants at the Minnesota Workshop, and three anonymous reviewers for extensive comments. Thanks also for support from the Social Sciences and Humanities Research Council of Canada (Standard Research Grant 410-2008-0400 to Ingo Brigandt) for supporting the Minnesota Center for Philosophy of Science’s Workshop on Integration in Contemporary
References (94)
Beyond reduction: Mechanisms, multifield integration and the unity of neuroscience
Studies in History and Philosophy of Biological and Biomedical Sciences
(2005)Relations among fields: Mendelian, cytological and molecular mechanisms
Studies in History Philosophy of Biological and Biomedical Science
(2005)- et al.
A genetic model for colorectal tumorigenesis
Cell
(1990) Integration of specialties: an institutional and organizational view
Studies in Biological and Biomedical Sciences.
(2013)- et al.
Toward transdisciplinary research: Historical and contemporary perspectives
American Journal of Preventative Medicine
(2008) Unifying science without reduction
Studies in the History and Philosophy of the Sciences Part A
(1977)When integration fails: Prokaryote phylogeny and the tree of life
Studies in History and Philosophy of Biological and Biomedical Sciences
(2013)On the use of the word, ‘epigenetic’
Current Biology
(2007)Scientific discovery and Maxwell’s kinetic theory
Philosophy of Science
(1987)Natural Obsessions
(1988)
The age distribution of cancer and a multistage theory of carcinogenesis
British Journal of Cancer
Cancer genetics: From Boveri and Mendel to microarrays
Nature Reviews Cancer
Integrating sciences by creating new disciplines: The case of cell biology
Biology and Philosophy
The Downs and ups of mechanistic research: Circadian rhythm research as an exemplar
Erkenntnis
Reduction, integration, and the unity of science: Natural, behavioral, and social sciences and the humanities
Towards an integrated view of neoplastic phenomena in cancer research
History and Philosophy of the Life Sciences
Hierarchies and causal relationships in interpretative models of the neoplastic process
History and Philosophy of the Life Sciences
Why don’t we get more cancer? A proposed role of the microenvironment in restraining cancer progression
Nature Medicine
Mina Bissell: Context is everything
Journal of Cell Biology
Beyond reduction and pluralism: Toward an epistemology of explanatory integration in biology
Erkenntnis
Explanation in biology: Reduction, pluralism, and explanatory aims
Science & Education
Expanding the social frame of knowledge: Interdisciplinary degree-granting fields in American colleges and universities, 1975–2000
Review of Higher Education
Inferring causation in epidemiology: Mechanisms, black boxes and contrasts
Mutation selection and the natural history of cancer
Nature
Cancer in the 20th century
The truth doesn’t explain much
American Philosophical Quarterly
The kinetics of papilloma formation in benzpyrene-treated mice
Cancer research
Explaining the brain: Mechanisms and the mosaic unity of neuroscience
Interfield theories
Philosophy of science
The importance of models in theorizing: A deflationary view
Philosophy of Science
The correlation between relatives on the supposition of mendelian inheritance
Philosophical Transactions of the Royal Society of Edinburgh
Dynamics of cancer: Incidence, inheritance and evolution
Problems of somatic mutation and cancer
BioEssays
Explanation and scientific understanding
The Journal of Philosohpy
Causal thinking and complex system approaches in epidemiology
International Journal of Epidemiology
Intratumor heterogeneity and branched evolution revealed by multiregion sequencing
NEJM
Ecological developmental biology
Conceptualizing the (dis)unity of science
Philosophy of Science
Model for the incidence of embryonal cancers: Application to retinoblastoma
Proceedings of the National Academy of Sciences of the United States of America
Improved endpoints for cancer immunotherapy trials
Journal of the National Cancer Institute
Distinct epigenetic changes in the stromal cells of breast cancers
Nature Genetics
Cited by (32)
Mechanistic inquiry and scientific pursuit: The case of visual processing
2022, Studies in History and Philosophy of ScienceCitation Excerpt :However, this feature of scientific inquiry is widely acknowledged within the philosophy of science. The principal idea that many different models may be fruitfully combined to help scientists make progress is familiar from debates about, e.g., explanatory perspectives (e.g., Giere, 2006; Potochnik, 2010), robustness (e.g., Wimsatt, 2007; Lehtinen, 2018; Ylikoski & Kourikoski, 2010), pluralism (e.g., Elgin, 2017; Mitchell, 2003), and explanatory integration (e.g. Brigandt, 2013a, 2013b; Kästner, 2018; O'Malley, 2013; Plutynski, 2013; Wayne, 2018).3 It is particularly nicely captured by mechanistic discovery accounts (Bechtel & Richardson, 2010; Craver & Darden, 2013).
On the nature of cross-disciplinary integration: A philosophical framework
2016, Studies in History and Philosophy of Science Part C :Studies in History and Philosophy of Biological and Biomedical SciencesCitation Excerpt :Integration has been a feature of the research agenda in the philosophy of science for decades, and it has taken many forms. In this literature, domain unification, for example, may be understood as a form of integration (e.g., Oppenheim & Putnam, 1958; but cf. Plutynski, 2013; Mitchell, 2002).12 Moving closer to the day-to-day business of science, talk of integration has blossomed in philosophy of biology.
Integration in biology: Philosophical perspectives on the dynamics of interdisciplinarity
2013, Studies in History and Philosophy of Science Part C :Studies in History and Philosophy of Biological and Biomedical SciencesArticulating Babel: An approach to cultural evolution
2013, Studies in History and Philosophy of Science Part C :Studies in History and Philosophy of Biological and Biomedical SciencesIntegrating data to acquire new knowledge: Three modes of integration in plant science
2013, Studies in History and Philosophy of Science Part C :Studies in History and Philosophy of Biological and Biomedical SciencesDimensions of integration in interdisciplinary explanations of the origin of evolutionary novelty
2013, Studies in History and Philosophy of Science Part C :Studies in History and Philosophy of Biological and Biomedical SciencesCitation Excerpt :Standards integration involves decisions about what kinds of data are necessary and whether the evidence is sufficient to support an explanation.2 Analyses of practices in microbiology, plant biology, and cancer research have converged on the significance of similar distinctions among types of integration for understanding how and why integration succeeds and fails in biology (Leonelli, this issue; O’Malley, this issue; Plutynski, this issue). We isolate and characterize three different levels of data integration that apply in the context of explaining the Cambrian Explosion and expose a lack of integration surrounding standards pertaining to causal explanations of the past.