Questions about anisotropy correction with a reference

[Guoqiang Huang]

Hi yuntao,

I am trying to alleviate structural elongation using the  'spisonet.py reconstruct' command. However, while the default parameters didn't significantly improve the structure, I observed a considerable improvement when I provided a reference map with the '--ref_resolution' set to 10 Å. The reference  map  was generated from an AF2-predicted model as follows: I fitted the high-confidence atomic model predicted by AF2 into the original map using ChimeraX and then generated a 10 Å low-pass filtered map of this model using the 'molmap' command. My questions are : 1) Is this reference map reasonable?  2) Is the corrected map reliable?

Thank you very much for your assistance.

Best regards,

Guoqiang Huang

[YUNTAO LIU]

Hi Guoqiang,

This is a great question whether it is legit to use low resolution alphafold structure to assist the spIsoNet training, and I do not have a definitive answer. 

I kind of understand your situation, and if we inject isotropic low resolution information into a map, with spIsoNet reconstruct, high resolution features may "grow" on that low resolution scaffold. 

spIsoNet with --ref_resolution strictly cuts off high resolution resolution from reference. I believe the high resolution information better than 10A actually comes from your data.

The problem is we do not know whether the alphafold structure is fully correct even at low resolution. Maybe alphafold is doing great with RMSD smaller than 1 or few Angstroms and maybe Alphafold is doing badly and fails to predict the structure. Currently probably no method can be used to verify independently. This is the reason that we are encouraged to obtain low resolution information through other approaches.

--
You received this message because you are subscribed to the Google Groups "spIsoNet" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spisonet+u...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/spisonet/131315d5-c2ac-462a-bf3a-1afa3bdea3c9n%40googlegroups.com.

--

Best Regards,

Yuntao Liu,  Postdoc.

California NanoSystem Institute

University of California Los Angeles