>>13402075>Using hybrid MD/DFT for the drug and MD for the rest seems clumsy to me. That may be, but it works quite well. Have a look at the paper i linked you, it outlines a very useful workflow which complimented the experimental work very well.
You are actually right in that MD doesn't necessarily yield good quantitative results, but does often produce very good qualitative ones. However, these qualitative results are what makes MD so useful. As opposed to experimental methods, we have access to the fully resolved system in MD, and as long as the qualitative behaviour is reproduced, we can gain real insight. It all has to be done with much care, as these are very often exceptionally complex models, with many layers of complexity built on top of each other. Particularly in my current research where I'm computing free energies of crystals. There's quite a lot of small little details and oddities to keep track of, and the risk of producing unphysical garbage is very real. But of course, similar issues may arise in experiment. At the end of the day, real insight has been gained from MD, and using it in conjunction with the experimentalists and theoreticians we may gain useful insight into how the world works.
