Hi Ethan, yes, I must be missing something indeed...
You are missing the point. What if you want to calculate Fcalc based on an existing model? What if you want to explore application of the _same_ TLS model to different sets of starting coordinates?
Yes, if I do need to calculate Fcalc based on existing model (that someone gives me) them I do need, indeed, these matrices. I just can't imginve a situation when I would ever do this this, but this is a separate question. Given model and data, and a tool to define optimal TLS groups - I could do all this myself, trivially.
And If I go to PDB I get all this information anyway from file headers.
What if you want to read in a TLS description that was generated through some other technique altogether, perhaps from trajectory analysis of MD or NMA?
That's a different story. Yes, in this case you are right of course, one would need to input T, L and S matrices explicitly. PHENIX allows you to do this:
/J. Appl. Cryst./ (2010). *43*, 669-676
Any time you want to compare two alternative models, you need first to be able to read them in.
Correct. But I typically verify the inputs by recalculating them, especially when it takes no effort to do it in this case.
It's all very well to say "the program knows best", but without the ability to compare what the program picks to what an external tool picks, how do you _know_ the program is "best".
I agree, but in this case you just need to use independent metrics to evaluate the end-result. Then you can safely by-pass all these "the program knows best" and will still on a safe side.
All the best! Pavel.
