Dynamic and thermodynamic models of adaptation

Adaptation through the lens of single-cell multi-omics data

“…To conclude, we are convinced that the models of adaptation, based on the application of "middle-out" ideas suggested in [1] must become a working tool for making sense of single-cell omics datasets. However, further work, both theoretical and related to practical data analysis, is needed to take into account the specificity of the data and the studied systems, which should lead to verifiable and reproducible conclusions.…”

“…Thirdly and quite fundamentally, the role of selection and inheritance can be much more important when studying adaptation to stress at the level of cells rather than whole organisms. Thus, cancer cells in culture are rapidly dividing at the time scale of experimental perturbation, which is not the case in the examples described in [1]. The adaptive changes are inheritable to some extent after cell division.…”

“…A.N.Gorban and his colleagues in their inspiring review described several theoretical models of adaptation and highlighted multiple examples convincing us in the existence of surprising at first thought phenomenon: the pattern of dynamical changes of basic statistical measures (correlation between features, their variance) can diagnose and prognose crises in the populations of objects exposed to stress [1]. The surprise is caused by the universality of this observation.…”

“…It looks very tempting to apply the thermodynamic models described in [1], to study the adaptation in single cell populations. Indeed, similarly to human beings, individual cells can possess variable levels of adaptation energy and resources: therefore, most of the assumptions look valid.…”

“…The first evident observation is that single-cell data are noisy and contain a large number of empty measures (so called technical drop-outs) [4]. These data are much noisier than the so-called "bulk" (using averages over millions of cells) measurements discussed in [1]. Various tricks such as using micro-pooling or variance stabilization are frequently used to improve the between-gene correlation structure in single-cell data, which, of course, impacts the quantification of the correlation graph.…”