Hi Luis,

I wrote the Fit in Map tool and fitmap command. It simply does a gradient ascent optimization of one of 3 metrics (overlap, correlation, correlation about mean) allowing a rigid rotation and translation in 3 dimensions. There is not much to this. It computes the metric gradient relative to these 6 degrees of freedom and takes a small step in that direction and continues iteratively until some convergence criteria is met. Advantages are it is fast (seconds) and robust (always converges). The main disadvantage is it just finds a local maximum of the metric which depends on the starting position. There often are many local maxima and it is not very good at finding them all, although it has a "search" option to try random placements. Other fitting programs like Situs will do a systematic search (rotate in steps of a certain number of degrees, translate in steps of some Angstroms) if you need that, can take hours. The Fit in Map approach has worked well despite this limitation because very often the researcher knows the approximately correct location of a molecule in a map, so it is senseless to do a global search that could take hours. Also for lower resolutions (> 10 A) or inaccurate atomic models there may be many seemingly equally good fits and a global search is likely to give wrong locations as the best fit. This is where the researcher applying their extra knowledge of the system and choosing the starting location helps -- for instance they know the antibody they are docking is on the surface of a molecular complex not buried in the middle. In general global searches don't take account of the many constraints the researcher knows and so they aren't often as helpful as you might think. Another aspect of Fit in Map is that the researcher develops understanding of the possible fits by doing them interactively. A global search is more objective but may not be as effective in developing the researcher's understanding of the ambiguities in the fitting. Basically the researcher has to pay more attention when they setup each local fit and it is much easier to ignore the subtleties when you instead get a long list of possible fits from a global optimization.

Tom

On Sep 3, 2020, at 10:06 AM, Elaine Meng meng@cgl.ucsf.edu wrote:

Hi Luis, I don't know of any specific discussion about the algorithm. However, it is a very commonly used tool (same method in Chimera and ChimeraX), mentioned by name in many high-quality publications, if that makes you feel any better. As I understand it, the local optimization just uses standard mathematical approaches. Best, Elaine

...

On Sep 3, 2020, at 9:55 AM, Luis José Castillo Valverde luis.castillo.valverde.98@gmail.com wrote:

Hello, I am currently working on my thesis work (this is why they have received many questions from me these days), this is why I am collecting information on biomolecular structures, and I wanted to know if you know of any article that comments on the pros and cons of its "Fit Map in Map" algorithm? This because I have been searching on this topic and have not been able to find anything. Sorry for the inconvenience and thanks in advance. Best regards, Luis

---

ChimeraX-users mailing list ChimeraX-users@cgl.ucsf.edu Manage subscription: https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users