Evening world! ( I have rewritten this 5 times at this point)

My visit to the BPS AM’19 was very short, but nevertheless I am happy that I had the opportunity to travel to this conference. Those few days have been eventful, and I have met and talked with multiple amazing people . Below I would like to share some of my thoughts on the talks, presentations and posters I have seen and engaged with. Since I have left early Monday morning, I cannot comment on the most of the content presented, so take this as a short snippet of my thoughts and not a comprehensive review.

Style

I firmly believe that style and presentation are extremely important in scientific world, especially during conferences. Good content organization and clear, simple to follow figures greatly improve audience’s ability to engage with material. I also think that organizing your content in a parsable way is the minimal courtesy you owe your readers.

I will start with some positive notes and remarks. First of all, black backgrounds in presentations are great! Of course there is a caveat of choosing appropriate color schemes for your figures (which I bet will deter people from going this route, since who wants to spend time making two sets of figures: for your presentation and for the paper), but when done well black background makes the content shine. The few presentations that had black background at the AM, warmed my heart.

Next, I need to applaud the usage of space on the slides. I haven’t seen a single overcrowded slide (maybe I just got lucky). Spreading out plots and figures so that they are readable is absolutely essential. Space usage is something I’ll come back to when talking about posters (including mine).

Now, let me move onto my incoherent rambling about style. I am in the process of learning Beamer, and it is definitely on the list of things I plan to master by the end of the year. It seems that the usage of LaTeX and LaTeX based tools and packages at this conference is almost exclusive to people coming in from the theory angle. The only talk I have seen for which slides were coming in the Beamer generated fashion was a theory talk on the inverse temperature dependence. I am not too surprised that this is the case, for two reasons. On one hand, majority of the CS, Math, Physics theory people would learn and use LaTeX anyway, which makes copy pasting and rearranging stuff easier. On the other hand, picking up LaTeX just for the presentations, especially if they are not peppered with formulas all the way through, seems like an overkill. I do hold a firm belief that people working in STEM fields should have some basic LaTeX skill as a part of the undergrad experience or early professional, but I would not go as far as to claim that we need to build everything in LaTeX. This is mostly an observation, as I reserve my LaTeX evangelism for a different post and time.

One thing that I do not grasp at all is why in 2019 we still have Comic Sans being used in presentations. Honestly, there is a plenty of great fonts that can catch people’s eyes if you are going for catchy. I do not share the ridiculous disgust towards Comic Sans, but even if we look at the original design specs it is supposed to be informal, children oriented font. Of course great research typed in Windings, will be ridiculous to read through, but that won’t change the fact that the underlying work is important. However, the font choices like this scream lack of professionalism and absolute style illiteracy to me. Just to reiterate this point, your style choices affect my perception of your attitude and skill, they do not affect my perception of your research work. Please, stop the Comic Sans abuse at science conferences!

Whew, okay, now let’s move to posters. First of all, while still far from perfect, I am fairly satisfied with the way my poster turned out. My main concerns are the crowded middle region where I had to have partially overlapping plots, and some not properly cleared up VMD renders that I simply did not have enough time to redo due to the data availability. Modulo those nuances I was fairly happy with the color scheme and organization, including some accidental design features. Coloring both the protein renders from VMD, and the lines in the plots with the same color convention was definitely a good move, as it allowed for seamless parsing of the plots, and saved much needed space by reducing number of legends I had to carry. An accidental feature was positioning the initial question with the illustrations at roughly same horizontal level as the conclusions. Since both contain a very similar illustration, it was easy to visually grasp what is that thing that changed over the course of our work. I am not fully pleased with some of the text in the poster, and I definitely would want to have another 2-3 days of re-reading and editing it, but alas.

The vast majority of the posters I saw at the conference were great. Clear organization, good posting sentences, and clear figures that stood out from the distance. However, there were a few common issues that I have noticed, predominantly in undergraduate posters. First is simple lack of content, or rather overabundance of white space. I have seen one too many posters with 2 or 3 figures, 3-4 paragraphs of text, and a canvas large enough to fit at least 3 more figures. I assume that this problem partially stems from the poster size expectation. Hey, we all think of 36″ x 24″ poster, and how this is a standard. However, I’d rather see more 24″ x 18″ posters that are content packed than 36″ x 24″ full of wide margins. Worst thing is that sometimes a poster with huge margins will still overcrowd the plots in the middle of itself. No bueno. Spacing out and arranging stuff is a pain in the butt, I know this myself. Plots come in weird proportions, plotting software is a monster to tame, and things tend to look different on your screen vs matte letter print vs glossy/matte poster print. However, the grind against just going with “Egh, good enough” is essential step in improvement of style. My work on this poster taught me a good principle, that my PI learned from his PI, so let me pass it onto you.

Make your poster, then print it out on Letter/A4 paper. If you can read the main figures clearly, then it’s a good font and line size, and proper spacing. If you are having trouble making sense of it on this print, go back to the drawing board.

NS paraphrasing Esmael Haddadian paraphrasing someone else…

On the other hand, I was impressed by some of the very creative approaches to poster presentations. Namely, the poster on Structure of a Non-canonical Middle Domain in Protocadherin-15 by Brandon L. Neel had an interesting twist to it. In addition to the regular static imagery, there also was a tablet set up, so that the interested people could watch MD trajectories evolve over time. I think that with the technology we have at hand right now, people who go extra mile to get their point across and present data in new and informative ways are the people with the higher potential for an impact.

On this note, I will wrap up my discussion of the style and presentation choices and do a quick dip into some of the research related moments and ideas.

Research

An obligatory, I am not a biophysicist by training. I am a computer scientist, I code things, analyze how fast they run, and think about how math models can help us solve real world problems. Thus, the rant below is not necessarily driven by deep professional concerns from a biophysics angle, but rather by some common sense scientific practices and beliefs.

I have attended some talks by scientists from D. E. Shaw research group, and while the content was impressive, I do feel deeply troubled by the messaging and approaches. Let me elaborate a little bit. Science by a large degree consists of trial and error. Especially repeated trial and repeated error. This applies to both experimental and simulation based results. If you were to measure or simulate something once, you would not publish a paper claiming it as a result. (Well sure in some rare cases you would, but that’s a more involved discussion.) Repeatability and multiple data points are essential to any kind of result that you are willing to cite as evidence for your hypothesis. Now, let us take this one step further. If I say that I did 700 independent experiments, and 95% of them yield a certain result with error margin of epsilon, you probably will agree that this sounds like a solid dataset. However, what if I also tell you that at this point in time no other research group can replicate my results? Sure, that also happens, sometimes the machinery is too complex and in many senses unique to the experiment. I mean after all both LHC and LIGO are not easy to come by at every other lab. On the other hand, those are large projects open to multiple collaborators within the field. They are not exactly a proprietary system with limited access given only to the employees of a particular private company. I think you see where I am going with this.

Work done by D. E. Shaw group is marvelous. However, I do have an issue with irreproducibility of the results. Say someone really wants to push for a certain drug affecting catalytic cycle of a protein involved in a disease. Assume that someone has ability to simulate the said catalytic cycle at high resolution with appropriate timescale. Now, what if someone accidentally messes up the parameters of their simulation? We get a perfect proof of effectiveness backed up by 500+ simulations, which then turns into active drug development. If 3 years down the line we discover that this in fact is bogus and doesn’t work, well we have wasted 3 years. If we don’t then we have a new drug, which might still end up being not quite as good as promised.

This is a long stretch hypothetical, but it has some ground behind it. Datasets have been cooked before (the example I am thinking of is a very particular test/train split in a study on the effects of certain drug combinations in cancer treatment), especially when money is on the line, and when pharmaceutical companies are in play, there always will be money on the line. Thus, while I hope that the researchers are keeping level heads and cold judgement, I cannot simply dismiss the concern about exclusivity of the technology that makes the MD research at D. E. Shaw beyond the cutting edge.

I believe that we should strive for repeatability and cross-validation in science. There always will be obstacles on this path, and I do believe in intellectual property rights, patents and related concepts. However, I think that the balance shall be maintained, and tools could and should be shared. After all making the first hammer is an invention, but once we have one it’s about who can do the most with it.

Concluding remarks

I started writing this almost a month ago, so I am not sure how connected those pieces feel. This was a long ramble about random stuff, so critique away at it.

Cheers, ns.