I struggle with good planning during conference meetings. We are only in a host city for a limited amount of time, and we are expected to not only engage in meaningful scientific discussion but also explore the surrounding community. This is my third time at a Biophysical Society Annual Meeting, and I do not yet have a solution.
As a simulationist, I of course find a good comparison between this problem and the one of the traveling salesperson. The traveling salesperson is the story of some unfortunate individual who, starting from home, is expected to go between a number of cities spaced out in a certain geometry and visit each city only once (hopefully making a sale) before returning home. This is directly relatable to our experience as conference attendees. Starting from the Ernest N. Morial Conference Center (or your hotel if you wish), you must not only work to your way to all posters, platform sessions, and symposia you’re interested in but also find your way to the most popular tourist destinations: Bourbon Street, Café du Monde, the Mercedes-Benz Superdome—how can you possibly manage this?
The problem presented by the traveling salesperson is one of note to computational scientists. It’s in the class of NP-complete problems, meaning that (in colloquial terms) this problem is very, very difficult and takes a lot of time to solve. Protein folding is another example of an NP-complete problem. To my students, I like to present a naïve (and classical) brute-force algorithm to fold proteins illustrating this point. You presume that you have a protein 100 amino acids in length, and you take the backbone phi and psi angles as the only interesting structural feature of this protein. Very ignorantly, you assume that each phi and psi angle can only adopt 3 possible conformations (a quite dramatic simplification). Excluding one phi and one psi angle from the termini, you can say that the protein therefore can adopt 3198 conformations. If your (somewhat old) computer takes 0.33×10-9 seconds to sample a single conformation, then you are expecting 2.8×1096 years of simulation time to go through all possible conformations. This is a very long time (much longer than the age of the current universe) and is therefore unthinkable to address computationally.
As biophysicists, we know that protein folding does not occur in a brute-force manner. Levinthal’s paradox, in a few words, stipulates that protein folding must be a directed and nonrandom process. In reality, proteins fold along a pathway driven by energetic and entropic demands. Likewise, there must be a way to direct our exploration of conformational space within the city of New Orleans—how can we minimize our travel along our trajectory through the city?
Despite the scientist in me wanting to construct an energetic function that considers my position and affinity to proximal points of interest, I know that I will not solve this problem during this year’s meeting. As the only alternative, I am more inclined to follow the fate of the randomly folding peptide, often at times making arbitrary choices out of sheer convenience. Yet, the desire within me remains to optimize my time and achieve true efficiency. Maybe next year.