The Latest in "Hands On" Software
Several years ago, the old 'molecular modeling software' industry collapsed due to competition from free, open-source, and frequently superior tools. The new paradigm is bringing us exciting improvements in functionality and diversity, provided one knows how to use it. The only thing I miss is the 'sandbox' interface tool just just plain old hacking around with molecules. Are there any new free tools emerging to take the place of old Behemoths like SYBYL and Discovery Studio?
There was a time, not that long ago, when the vast majority of computational drug discovery research professionals spent a substantial amount of their time staring at one very specific form of software window -- a molecular modeling interface. Ten years ago, there was a small handful of successful corporate solutions (e.g., Tripos, Accelrys, Schrodinger to name some of the largest) that dominated the market. The basic upshot of all of this was that regardless of whether your research was in academic, industry or government settings, there was a fair chance that you used one of these well-established tools, and paid a fair bit of money each year to do so.
Wow -- things have certainly changed! As of April 2015, if I do a Google search on "molecular modeling software" only one out of the top ten hits (HyperChem) is a commercial product. The rest of the top links are dominated either by freely distributed programs or else by lists of different resources available. What happened to the old software structure that once so dominated the computational support of medicinal chemistry research?
The precise reason for the collapse of small but once-vibrant commercial sector is difficult to pin down, but it is probably fairly telling that the old modeling environments are generally not missed. An increasingly sophisticated user base has grown more comfortable in fusing a collection of different, separately-distributed tools together in order to carry out their simulations and analysis. Thus a researcher might now use Marvin to sketch and structurally optimize a chemical's geometry, Autodock to predict how it might bind to a target receptor, and PyMol to produce publication-quality graphics of the predicted complex. The researcher has to flip between windows of the workstation, and it may feel a bit more cumbersome than the single unified environment that we once knew, but the actual calculations are typically a lot faster, and probably more reliable.
As a devout proponent and user of open source scientific software (see my Tumblr blog for proof), it would be hard to find fault with the current situation, which I believe is far more indicative of a vibrant scientific culture than the old black box era. If there's been anything that I've missed about the monolithic tools such as SYBYL, Jaguar and Discovery Studio though, it has been the facile opportunity to just goof around. By this I mean, be able to rapidly assemble and tweak a molecule, optimize it, run quick dynamics simulations to understand conformational variants, rapidly ascertain how minor functional group variations affect properties, and so forth. In other words, a computational and quantitative equivalent to an elder generation's use of solid ball and stick models. The ability to generate nice graphics to characterize interesting features of the molecule is, of course, very helpful as well.
A question that I've asked myself from time to time over recent years is whether a free (and ideally open-source) tool will emerge that serves this purpose. Below is a graphic (admittedly an unscientific kludge) that I hacked together based on my current opinions of some great but not completely satisfactory software options. The areas of scoring are pretty straightforward:
- Builder: can you build an organic molecule or peptide without too much hassle?
- Optimizer: can you take the hand-built molecule and convert it into a physically plausible geometry with little to no teeth-gnashing?
- Dynamics: can you see what soft of conformations that molecule is likely to sample under different conditions of temperature and dielectric medium?
- Graphics: can you generate and capture some nice pictures of your molecule?
|Avogadro (version 1)||5||3||0||2|
The unscientific scoring scheme I've used basically ranges from '0' (the software doesn't support this) up to '5' (in my opinion, the program does this better than any of the old commercial products once did). For reference, I probably would have given a program like SYBYL 8.1 scores of 3 and 4 across the board. As of right now, I don't think any of these free tools are quite to this level, although clearly some tools do exceed prior standard in some areas. I absolutely love Avogadro's small molecular and peptide builders (version 1 only -- there are things in version 2 that are sadly very broken!), and I use Vega-ZZ extensively, crunching out simple molecular mechanics calculations on covalent inhibitors because the program is so fast and easy). PyMol, of course, is practically an industry standard now for high quality graphics. So sad that Warren Delano didn't survive to see his legacy -- the man truly deserves to be a millionaire on a beach somewhere!
If you examine the collection as a whole (i.e., overlook the fact that each of these tools has strengths and weakness, and none is quite the 'perfect' package) you'll see that the one weakness is that none of these tools does a great job of handling molecular dynamics. My complaint, which is articulated a lot more verbosely elsewhere in this blog post, is that while three of the tools support some level of interactive dynamics monitoring, none of them make it anywhere as simple as the methodologically analogous molecular mechanics calculations. In all cases, you really have to use a separate tool such as VMD to parametrize the system, or else supposedly simpler formulations (e.g., representing the system with highly generalized CVFF or Tripos force fields) invariably fail without a helpful explanation. Perhaps what these software developers don't realize is that there's a substantial demand, not just for highly rigorous dynamics simulations that can be validated as plausibly reproducing accurate IR spectra or kinetics measurements), but also for those simple sandbox glimpses that answer intuitive questions such as, 'I wonder if this molecule might ever vibrate such that atom A gets close to atom B?' I used to use SYBYL and Insight II for goofy things like that all the time. Now it seems like I'm stuck doing a semi-publication-quality NAMD simulation every time I want to watch my molecule wobble around.
Ah well -- it's wonderful and amazing what one can get for free these days. I guess I can put up with a little less wobbling.