LiveCoMS “Best Practices” articles: Put your simulations on the shoulders of giants

2018-08-20

In a now-classic New Yorker article, and later in his book, Atul Gawande discussed how surprisingly simple checklists can drastically improve medical treatment. For example, a simple six-point checklist list for reducing the occurance of central line infections has saved thousands of lives and saved tens of millions of dollars. As Gawande also points out, the incredible safety record of commercial aviation, with no commerical airline deaths worldwide in 2017 and an overall estimated fataility rate of one per 16 million flights, is in large part due to the extensive preflight and emergency checklists that pilots use.

We believe that molecular simulation can benefit from development of such checklists. For many years, high computational cost was the biggest reason for why simulation couldn’t accurately model complex molecular systems. As computers have become orders of magnitude faster over the last several decades, however, other issues have become bottlenecks in performing useful molecular simulations in a growing number of cases. For example:

  • It takes significant expertise to generate usable results.
  • It is difficult to validate simulations relative to previous computations.
  • There is lack of reproducibility across the available simulation codes.

It’s hard to think of a situation where someone’s immediate life is on the line when a simulation fails to conserve energy or a protein heats to 8 million degrees. But there is a real cost to bad simulations, both in researcher time and in lost opportunities for progress. The field stagnates if we continually need to repeat simulations again and again in our research groups, or deal with bad data in the literature. So, if we know molecular simulations have a high failure rate, can we identify simple steps that can improve this? Can we build on the lessons that other people have learned? In other words, where are the checklists for molecular simulation?

The Best Practices article category in LiveCoMS is designed for papers that present checklists and, importantly, the detailed thinking going on behind these recommendations for certain calculations or family of calculations. Since the best practices change as algorithms improve, computational limitations relax, and science progresses, these papers, like other LiveCoMS papers, are intended to be updated regularly. A living journal approach is especially vital for best practices. Because the field is moving so fast, these changes often take place on the 1–3 year time scale, requiring corresponding response times. These documents are also intended to reflect not just a single opinion, but broad consensus across several research groups on what the best practices in a given subject are, and what steps researchers can take to minimize the chance of failure to carry out the intended computational task.

We imagine that these articles will become an invaluable resource to teach researchers how to perform simulations the right way, raising the bar for publication-quality science. Standing on the shoulders of those that came before, to paraphrase Isaac Newton, will help us see that much farther. We can devote more of our time to answering our driving questions, rather than redoing work. So we look forward to your help in developing best practices for your community, and seeing them appear in LiveCoMS!