It is important to point out that a model need not necessarily be
right in order to be useful. In fact, every model is only an approximation and abstraction of the biological system under study and
will be replaced by better approximations when additional information becomes available. “All models are wrong, but some are useful” [33]. Even an eventually falsified model may suggest useful
experimental studies that would otherwise not have been performed, and thus help in advancing our knowledge of a biological
system. Such models should receive deserved credit.
2.5 Perform
Corresponding
Experimental Studies
As Katchalsky pointed out [24], “Theory tells us what cannot
happen, and it can tell us what could happen. But only experiments
tell us what does happen.” All model predictions need to be subject
to subsequent experimental tests.
2.6 Go Back to Step
2 and Iterate;
Expansion of Model
(Even After Publishing
the Original Work)
It has become more and more common to see studies that have
iterations between modeling and experiments. Sometimes the
integrated experiment-modeling process may even lead to revisiting
step 1 to define new questions and seek expanded collaborations.
For example, early modeling studies [9, 13] on EMT focused on
the core regulatory network (Fig. 1). Several subsequent studies
expanded the network to explore how additional factors contribute
to the spectrum of EMT phenotypes [4, 34, 35].
3 Modeling Population Heterogeneity in EMT
Intra-tumoral heterogeneity, wherein cancer cells within the same
tumor exhibit different phenotypes, has been reported across multiple cancer types, both in vitro and in vivo [36]. Tumor cell
populations in different cancer types including leukemia [37],
breast cancer [38], colorectal cancer [39, 40], brain cancer [41],
and prostate cancer [42] can consist of subpopulations of cells that
exhibit stem-cell-like behavior. Cells in triple-negative breast cancer
can exhibit distinct phenotypes including luminal, basal, immunomodulatory, mesenchymal, and stem-like [43]. In small cell lung
cancer, tumor cells can exhibit both neuroendocrine and
non-neuroendocrine phenotypes [44]. Intra-tumoral heterogeneity has recently been identified as a principal cause for the failure of
anticancer therapies [45]. Therefore, characterization of the
mechanisms driving this feature of tumor cell populations is key
to advancing anticancer therapeutics. In many (perhaps most)
cases, genetic heterogeneity does not underlie phenotypic heterogeneity, i.e., tumor cells exhibit different phenotypes in spite of
carrying the same genetic alterations. This indicates that nongenetic mechanisms may be the chief driver of intra-tumoral
heterogeneity.
Mathematical Modelling of EMT
391
right in order to be useful. In fact, every model is only an approximation and abstraction of the biological system under study and
will be replaced by better approximations when additional information becomes available. “All models are wrong, but some are useful” [33]. Even an eventually falsified model may suggest useful
experimental studies that would otherwise not have been performed, and thus help in advancing our knowledge of a biological
system. Such models should receive deserved credit.
2.5 Perform
Corresponding
Experimental Studies
As Katchalsky pointed out [24], “Theory tells us what cannot
happen, and it can tell us what could happen. But only experiments
tell us what does happen.” All model predictions need to be subject
to subsequent experimental tests.
2.6 Go Back to Step
2 and Iterate;
Expansion of Model
(Even After Publishing
the Original Work)
It has become more and more common to see studies that have
iterations between modeling and experiments. Sometimes the
integrated experiment-modeling process may even lead to revisiting
step 1 to define new questions and seek expanded collaborations.
For example, early modeling studies [9, 13] on EMT focused on
the core regulatory network (Fig. 1). Several subsequent studies
expanded the network to explore how additional factors contribute
to the spectrum of EMT phenotypes [4, 34, 35].
3 Modeling Population Heterogeneity in EMT
Intra-tumoral heterogeneity, wherein cancer cells within the same
tumor exhibit different phenotypes, has been reported across multiple cancer types, both in vitro and in vivo [36]. Tumor cell
populations in different cancer types including leukemia [37],
breast cancer [38], colorectal cancer [39, 40], brain cancer [41],
and prostate cancer [42] can consist of subpopulations of cells that
exhibit stem-cell-like behavior. Cells in triple-negative breast cancer
can exhibit distinct phenotypes including luminal, basal, immunomodulatory, mesenchymal, and stem-like [43]. In small cell lung
cancer, tumor cells can exhibit both neuroendocrine and
non-neuroendocrine phenotypes [44]. Intra-tumoral heterogeneity has recently been identified as a principal cause for the failure of
anticancer therapies [45]. Therefore, characterization of the
mechanisms driving this feature of tumor cell populations is key
to advancing anticancer therapeutics. In many (perhaps most)
cases, genetic heterogeneity does not underlie phenotypic heterogeneity, i.e., tumor cells exhibit different phenotypes in spite of
carrying the same genetic alterations. This indicates that nongenetic mechanisms may be the chief driver of intra-tumoral
heterogeneity.
Mathematical Modelling of EMT
391