by Fred Ross Last updated: June 27, 2009

My coauthor and I received a rejection from PLoS Computational Biology with the comment, "it is an interesting result but more theoretical than computational and more mathematical than biological despite its possible relevance to biological processes." This editor's confusion is remarkable.

Computational physics is a branch of theoretical physics. Computational chemistry is a branch of theoretical chemistry. They are subfields where theoretical questions are answered with tools executed on a computer. Both still require careful analytical work like any other theorist to establish what to compute and the precision and accuracy of the computation. The only definition I can conceive which would separate computational science from theoretical is "the use of black boxes to generate results with no understanding of how they work" -- and if this be the case, let us end computational biology today!

Why do computational biologists avoid the label of "theory"? In part because of a sour grapes reaction among many biologists to the successes of theoretical physics, and in part because of the rise of that muddled label "*in silico*".

The ubiquitous tags *In vivo* and *in vitro* are holdovers from when biologists actually used Latin and Greek. *In vivo* describes a process occurring in the organism where it is usually found, *in vitro* the reconstitution of the process in a test tube. Both are clear so long as the process, be it glycolosis or the growth of a pathogen, is clearly stated.

*In silico* conflates simulation and data analysis. Anyone who has done both will tell you that the two activities, though they both involve mathematics, are otherwise completely distinct.

Data analysis is directly connected to reality. What the data represents is not in question. We plot, fit, plot again, measure and test, but the tools are built to elucidate a structure we already know to be relevant.

Simulation begins in a vacuum, and must be forcibly attached to reality. We start with some mathematical description of a phenomenon, and steadily approximate it, fix numerical values in it, and massage it until it is reduced to a problem the computer can solve. The whole path along the way must be documented and justified. We are trying to demonstrate a known structure.

Let us lay that misused Latin tag to rest.

The second reason, "more mathematical than biological", is simply puzzling. We began from a biological problem, we ended with an application of our solution. Inbetween came rigorous mathematics, but this could only make the study abiological if biology were intrinsically immiscible with mathematics. If our mathematics had been chemistry, would we have magically been doing biology?

We have naturally taken the paper elsewhere.

Fred Ross

June 2009

Lausanne, Switzerland