PgmNr P2020: How a Framework for Evolutionary Systems Biology Can Accelerate Reproducible Modeling of Mechanistic Fitness Landscapes.

Authors:
Laurence Loewe


Institutes
Laboratory of Genetics and Wisconsin Institute of Discovery, University of Wisconsin-Madison, Madison, WI.


Abstract:

Systems approaches are becoming increasingly important, as  population genomics aims to understand factors shaping the reproduction and survival of genotypes based on the phenotype they encode and their environment.  Current interest in personalized medicine, disease prediction and understanding GxPxE interactions can easily be frustrated by the computational work required and by the limitations of current tools, which often excel at the specialized tasks they were written for, but are of little use for closely related tasks that often only differ in minor details. Considering the computational challenges of mechanistically predicting phenotypes, diseases or fitness  in non-trivial systems raises the question, how often it will be possible to extract reliable biology from simulations in which signals can be hard to separate form the noise of mishandling data, numerical errors, bugs, logic errors and more. Sound programming and modeling advice emphasize the importance of good questions to solve specific problems. This often leads to results or tools that are highly specific to some type of question, ignoring other closely related questions. Asking such questions often triggers so much code-wrestling, data shoveling, version wrangling and other silicon digging, that little time remains for considering big picture questions, such as how the relentless New Evolutionary Synthesis that started in the 1920s can be advanced to the point where fitness landscapes become useful in the real world. 

To facilitate the development of computational tools for ultimately enabling  interactive "flight simulators for mechanistic fitness landscapes", it is pivotal to understand the corresponding requirements and to provide a sound and usable conceptual framework that is stable enough to serve as a solid basis for writing long-term backwards compatible code that is reproducible enough to enable efficient model-reuse, so researchers can focus again on the biology instead of resolving tedious computational problems.

Earlier work, discussions at a series of EvoSysBio meetings, and newer advances enabled the development of such a framework for Evolutionary Systems Biology that combines the rigor and expertise of molecular biology and other Intra-Organism Biology results with the rigor of models in Population-Genetics Biology and other Trans-Organism Biology. The result is a redefinition of EvoSysBio that lays out in much improved clarity and precision the long road to enabling fully automated mechanistic predictions of phenotypes and diseases from genotypic information. Cancer cell evolution will used as an example.