Hear, hear! Although I like the Latin expression and the occasional rant in general, I fully agree with Gerlind's point.
TLS refinement is a great tool and it can help even if the model segmentation is not biologically relevant. However, not biologically relevant may not necessarily mean that the model is overfitted. This is what we test in PDB-REDO: we take all defined TLS models (typically the model in the coordinate file and one-group-per-chain, but you can specify mode models) and refine with these. Then we use a version of the Hamilton test to see if the improved fit with the data comes from a genuine improvement of the model rather than from just adding extra parameters. This approach has flagged quite a few cases in the PDB where the TLS segmentation is a bit "overenthusiastic". However in the case of 8q66 the more complex model is, from a refinement point-of-view, fine. I haven't read the paper for this structure model, but if the authors are not deriving any specific conclusions from the TLS model, I do not see it as a problem. As Gerlind points out, we should teach the right things. In this case there are a few things users should know: - TLS refinement is often a means to an end, not a (biological) model analysis. - TLS analysis (i.e. grouping refined (anisotropic) B-factors to fit a domain-like model of "motion") on the other hand is a useful tool if done carefully. - Do not read too much into a given TLS model if you do not know where it came from. Also don't if you do. - Don't overparameterise your TLS model, apply Occam's razor, and test for overfitting. - You can check your TLS model with tools like Parvati. Perhaps having more validation tools would be nice. Cheers, Robbie > -----Original Message----- > From: CCP4 bulletin board <CCP4BB@JISCMAIL.AC.UK> On Behalf Of Gerlind > Sulzenbacher > Sent: Tuesday, October 29, 2024 17:08 > To: CCP4BB@JISCMAIL.AC.UK > Subject: Re: [ccp4bb] Structures refined ad mentula canis (this is Latin) > > Dear Italo, > > It seems to me that the ccp4bb has always been a very gentle and empathic > platform of exchange of ideas and offering a great deal of help to non- > experts, "ignorants" as your friend calls them. > > I remember that in the late 90thies there were some ayatollahs, which have > become gentle civils by now, that used accusatory citations to underpin their > claims. This period is over, or at least I hope so. > > > If ignorance exists, it's our fault of not being able to teach properly the > young > generation. > > I understand your frustation and may be it deserves a letter in an Acta > Journal, but please, please, refrain of citing the "ignorants" in this bb. > > Wishing you a happy retirement and nice and fruitfuil projects, other than B- > factors, for your future. > > Gerlind > > > > > > On 29/10/2024 16:38, Italo Carugo Oliviero wrote: > > > Ce mail provient de l'extérieur, restons vigilants > > In a few months, I will retire, because my brain thinks it is still > young > but the rest of the body disagrees and thinks the brain is slightly senile. Of > course, I will miss the protein structures very much, so unpredictable in > their > insolent beauty. > > > > > I have devoted several years of my life to the B-factors of their atoms. > > > > > Initially, almost 30 years ago, some colleagues mocked me, telling me > that the B-side of proteins was nothing but dust hidden under the carpet, > nothing but background noise. Some later changed their minds, and it is now > generally agreed that the B-side of proteins deserves to be visited and > admired. > > > > > However, I am concerned about more recent developments that are > turning B-factors into real junk. TLS refinements as audacious as they are > ridiculously unprofessional are producing disasters. For example, the 8q66 > structure is very strange. It has been split into individual TLS mini-blocks, > which clearly demonstrates that the authors are unaware of the TLS > procedure and its proper application. A protein containing two structural > domains was divided into six small TLS blocks, one of which belongs to both > domains. Where are physics and chemistry? > > > > How can this happen? Here is the answer from a friend and > colleague: “unfortunately, the crystallographic programs became so > automatic and easy to use nowadays, that even complete ignorant people > can get something from them, and do not think if what they got makes sense > or not.” > > > > > > > Are computational models preferable in the end? > > > ________________________________ > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA- > JISC.exe?SUBED1=CCP4BB&A=1 > > > > > -- > Gerlind Sulzenbacher > Architecture et Fonction des Macromolécules Biologiques > UMR7257 CNRS, Aix-Marseille Université > Case 932 > 163 Avenue de Luminy > 13288 Marseille cedex 9 > France > Tel +33 413 94 95 27 > E-mail: gerlind.sulzenbac...@univ-amu.fr > <mailto:gerlind.sulzenbac...@univ-amu.fr> > > ________________________________ > > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
