| Thank you for having this conversion. I have heard this question more than once (why is experimental determination needed when there is Alphafold) at grant study sections. Some came from structural biologists, so hopefully, the discussion here would help them also. Computational modeling had existed for decades before Alphafold, and no one ever questioned then, as far as I am aware, the need for experimental determination of the modeled structures regardless of how accurate the predicted models may have been (for example an MD state of a point mutation using a high-resolution X-ray model as a starting model), so I am confused as to why we are asking this question now. In my view, predicted model, as good as it may be, is still a prediction.
Cheers, Quyen Quyen Hoang, PhD Associate Professor Department of Biochemistry and Molecular Biology Adjunct Associate Professor of Neurology Principal Investigator of the Stark Neurosciences Research Institute Indiana University School of Medicine 635 Barnhill Drive Medical Sciences Building, room MS0013C Indianapolis, IN 46202
On Apr 2, 2023, at 10:52 AM, Eugene Valkov <eugene.val...@gmail.com> wrote:
It depends on the problem under investigation. For example, if you propose to perform drug discovery with membrane proteins in a lipid/detergent environment or study the structural organization of large ribonucleoprotein assemblies, then AlphaFold will be of limited value and you should point this out in the rebuttal. If one of your aims is to study a protein, divide it into pieces, and solve structures of domains, then this is justifiably questioned by reviewers in light of the information provided by structure prediction. Perhaps in the reviewer's opinion, there is no need to use substantial funds and time to solve that structure. A more competitive proposal may then re-focus on requesting resources for using the structural hypotheses to probe the mechanism with more ambitious experimental structural aims or to take a deeper dive into the function. As Darwin said, it is not the strongest or the fastest (or even the most intellectual!) that survives.
That’s the point. All these predictions can not replace experiments and they should be used only in the absence of experimental structures, that too with caution. A biophysicist and structural biologist understand this but most of the non-experts (including
the reviewers of the grants from non-structural biology study sections) don’t understand this and think that with alphafold, there is no need for experimentally determined structures. That’s more damaging than helpful.
Predicted
structures lack the precision and accuracy of experimentally-determined structures at high resolution with all the benefits of unbiased co-discovery of solvent molecules, ions, ligands, etc., bound to molecules of interest.
It
is also true that AI-assisted structure prediction can be stunningly accurate to the level of correctly identifying interfacing residues between interacting proteins and accurately recapitulating the architectures of large, multi-domain complexes. AI-assisted
structure prediction is immensely liberating in lowering the threshold to generate testable hypotheses for those lacking access to experimental structural biology resources.
AlphaFold
and related structure prediction tools are here to stay, and no magic incantations in funding proposals will likely make them disappear. So rather than wring our hands as a community and mourning the loss of ‘divide-and-conquer’ structural biology, which was
a rich vein to mine over the last decades, we should adapt to structure prediction and normalize its use, with proper controls and caveats, as part of our arsenal of tools and methods to focus on the questions under investigation.
With
best wishes,
Eugene
Eugene
Valkov, D.Phil.
Stadtman
Investigator
RNA
Biology Laboratory
Center
for Cancer Research
National
Cancer Institute
Frederick
MD 21702, USA
(301)
846-1823
I am also not sure whether AlphaFold can address the impact of ions and other cofactors on the fold of many proteins.
Best wishes,
Jacinto
On 2/4/23 16:20, Srivastava, Dhiraj wrote:
May be this article is of some help suggesting the need of experimental structures despite excellent alphafold model.
All, the first hurdle will of course be whether the AlphaFold model works as a MR model, even with the 100% completeness and sequence identity of a bespoke model. The question is what B factors to use or which disordered
bits to leave out, as that can strongly influence the result (perhaps use info from a similar structure). If it doesn't work in MR that's a pretty good indication that it's too far from reality to be useful for looking at detailed interactions.
Does anyone know of a systematic investigation of the success rate of AlphaFold models in MR ? That would be useful ammunition !
Cheers
-- Ian
Dear all,
I think that quoting general viewpoints and statements, however
knowledgeable and respected their authors may be, will only exacerbate the
climate of clashing prejudices between two camps and is bound to sustain a
war of opinions rather than lead to a rational acceptance that something has
changed. The frustration is that one camp (the AlphaFold believers) can be
viewed as in effect preventing experiments that could prove it wrong.
One way to deal with this obstruction would be to provide, in each
particular case, evidence that the AlphaFold results "do not cut it" as the
sole provider of 3D information within the project at hand. This means that
every grant proposal requesting resources towards a crystallographic
structure solution should document the fact that AlphaFold predictions have
been performed (or, often, looked up in a database of pre-cooked results)
but do not provide the accuracy required for the proposed investigation. If
this step of writing up the "Background" section of the grant actually
delivers a useful result, then everyone will be happy; and if it doesn't,
then the case for the need to allocate resources to solving the structure by
crystallography will be unassailable. In this way, AlphaFold will be a game
changer (we have known that since July 2021) but not a game killer. Savvas
alluded to a similar approach, but it could be made a formal requirement
acceptable to both proposers and reviewers, who would then both be dealing
with the situation in a scientific rather than dogmatic manner.
With best wishes,
Gerard.
--
On Sat, Apr 01, 2023 at 06:54:33PM +0000, Goldman, Adrian wrote:
> I think this is all true - and I’ve been putting things like this into my (failing) grants - but I get the dispiriting sense that the medics think (to borrow a line from hamlet) “the applicant doth protest too much methinks”.
>
> Well if as per James H today ;), we deposit coordinates to 1sf, alphafold will be just fine.
>
> Of course the coordinates won’t be of any use to anybody, but the pictures will be nice.
>
> Adrian
>
> Sent from my iPhone
>
> > On 1 Apr 2023, at 21:39, Randy John Read <rj...@cam.ac.uk> wrote:
> >
> > There’s also this preprint with Tom Terwilliger as lead author:
https://www.biorxiv.org/content/10.1101/2022.11.21.517405v1. The title is “AlphaFold predictions: great hypotheses but no match for experiment”.
> >
> > Best wishes,
> >
> > Randy
> >
> >> On 1 Apr 2023, at 18:18, Savvas Savvides <00009d24f7f13e09-dmarc-requ...@jiscmail.ac.uk> wrote:
> >>
> >> Dear Rams,
> >>
> >> I salute you for sharing this.
> >>
> >> Just a week ago, I also received a remark along these lines on a declined grant application. The remark was the only unfavourable point, which suggested that it must have weighed disproportionally towards the negative outcome. This was a two-stage evaluation
process and the grant was cut in stage-1 where it was evaluated by a small group of evaluators, none of whom was a structural biologist/biochemist. Stage-2 would have involved peer review by international experts.
> >>
> >> Despite my initial disbelief about what this remark might have caused and upon reflection, I realized that it might be time to become proactive in future applications in anticipation of the apparent growing trend towards such remarks and perceptions.
> >>
> >> I think that a generalized form of preemptive text might not serve the purpose well, but perhaps well-articulated statements specific to the proposed biological problem at hand (perhaps aided by illustrations demonstrating the inability of structure prediction
to address the problem at hand) might be the better way to go. Even though many of us who teach courses in experimental structural biology and structural bioinformatics at undergraduate and graduate levels are already actively addressing many of these issues,
there is a much bigger and far more senior scientific population out there that makes important decisions on science policy/funding/infrastructures/evaluations/recruitment/etc that are not getting such educational exposure.
> >>
> >> The following resources provide good material and starting points to reflect and elaborate upon.
> >>
> >> The article by Perrakis and Sixma in EMBO Reports
https://www.embopress.org/doi/full/10.15252/embr.202154046
> >>
> >> The recent comment paper in Nature Methods by Thomas Jane
> >>
https://doi.org/10.1038/s41592-022-01760-4
> >>
> >> A correspondence in Science by Moore, Hendrickson, Henderson and Brunger
> >>
https://www.science.org/doi/10.1126/science.abn9422?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%20%200pubmed
> >>
> >>
> >> Best wishes
> >> Savvas
> >>
> >>
> >> ----
> >> Savvas Savvides
> >> VIB Center for Inflammation Research
> >> Ghent University, Dept. of Biochemistry & Microbiology
> >> Technologiepark 71, 9052 Ghent, Belgium
> >>
> >> Email: savvas.savvi...@ugent.be ;
savvas.savvi...@irc.vib-ugent.be
> >> Phone: +32 (0)472 928 519 (mobile) ; +32 (0)9 331 36 60 (office)
> >> Web:
https://savvideslab.sites.vib.be/en#/
> >>
> >>>> On 1 Apr 2023, at 16:57, Subramanian, Ramaswamy <subra...@purdue.edu> wrote:
> >>>
> >>> Ian,
> >>>
> >>> Thank you. This is not an April fools..
> >>> Rams
> >>> subra...@purdue.edu
> >>>
> >>>
> >>>
> >>>> On Apr 1, 2023, at 10:46 AM, Ian Tickle <ianj...@gmail.com> wrote:
> >>>>
> >>>> ---- External Email: Use caution with attachments, links, or sharing data ----
> >>>>
> >>>>
> >>>> Hi Ramaswamy
> >>>>
> >>>> I assume this is an April Fool's but it's still a serious question because many reviewers who are not crystallographers or electron microscopists may not fully appreciate the difference currently between the precision of structures obtained by experimental
and predictive methods, though the latter are certainly catching up. The answer of course lies in the mean co-ordinate precision, related to the map resolution.
> >>>>
> >>>> Quoting
https://people.cryst.bbk.ac.uk/~ubcg05m/precgrant.html :
> >>>>
> >>>> "The accuracy and precision required of an experimentally determined model of a macromolecule depends on the biological questions being asked of the structure. Questions involving the overall fold of a protein, or its topological similarity to other
proteins, can be answered by structures of fairly low precision such as those obtained from very low resolution X-ray crystal diffraction data [or AlphaFold]. Questions involving reaction mechanisms require much greater accuracy and precision as obtained
from well-refined, high-resolution X-ray structures, including proper statistical analyses of the standard uncertainties (s.u.'s) of atomic positions and bond lengths.".
> >>>>
> >>>> According to
https://www.nature.com/articles/s41586-021-03819-2 :
> >>>>
> >>>> The accuracy of AlphaFold structures at the time of writing (2021) was around 1.0 Ang. RMSD for main-chain and 1.5 Ang. RMSD for side-chain atoms and probably hasn't changed much since. This is described as "highly accurate"; however this only means
that AlphaFold's accuracy is much higher in comparison with other prediction methods, not in comparison with experimental methods. Also note that AlphaFold's accuracy is estimated by comparison with the X-ray structure which remains the "gold standard"; there's
no way (AFAIK) of independently assessing AlphaFold's accuracy or precision.
> >>>>
> >>>> Quoting
https://scripts.iucr.org/cgi-bin/paper?S0907444998012645 :
> >>>>
> >>>> "Data of 0.94 A resolution for the 237-residue protein concanavalin A are used in unrestrained and restrained full-matrix inversions to provide standard uncertainties sigma(r) for positions and sigma(l) for bond lengths. sigma(r) is as small as 0.01
A for atoms with low Debye B values but increases strongly with B."
> >>>>
> >>>> There's a yawning gap between 1.0 - 1.5 Ang. and 0.01 Ang.! Perhaps AlphaFold structures should be deposited using James Holton's new PDB format (now that is an April Fool's !).
> >>>>
> >>>> One final suggestion for a reference in your grant application:
https://www.biorxiv.org/content/10.1101/2022.03.08.483439v2 .
> >>>>
> >>>> Cheers
> >>>>
> >>>> -- Ian
> >>>>
> >>>>
> >>>> On Sat, 1 Apr 2023 at 13:06, Subramanian, Ramaswamy <subra...@purdue.edu> wrote:
> >>>> Dear All,
> >>>>
> >>>> I am unsure if all other groups will get it - but I am sure this group will understand the frustration.
> >>>>
> >>>> My NIH grant did not get funded. A few genuine comments - they make excellent sense. We will fix that.
> >>>>
> >>>> One major comment is, “Structures can be predicted by alpfafold and other software accurately, so the effort put on the grant to get structures by X-ray crystallography/cryo-EM is not justified.”
> >>>>
> >>>> The problem is when a company with billions of $$s develops a method and blasts it everywhere - the message is so pervasive…
> >>>>
> >>>> Question: Is there a canned consensus paragraph that one can add with references to grants with structural biology (especially if the review group is not a structural biology group) to say why the most modern structure prediction programs are not a substitute
for structural work?
> >>>>
> >>>> Thanks.
> >>>>
> >>>>
> >>>> Rams
> >>>> subra...@purdue.edu
> >>>>
> >>>>
> >>>>
> >>>>
> >>>> 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
> >>>
> >>
> >>
> >> To unsubscribe from the CCP4BB list, click the following link:
> >>
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
> >>
> >
> > -----
> > Randy J. Read
> > Department of Haematology, University of Cambridge
> > Cambridge Institute for Medical Research Tel: +44 1223 336500
> > The Keith Peters Building
> > Hills Road E-mail:
rj...@cam.ac.uk
> > Cambridge CB2 0XY, U.K.
www-structmed.cimr.cam.ac.uk
> >
> >
> > ########################################################################
> >
> > 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/
>
> ########################################################################
>
> 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/
########################################################################
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/
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
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
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
|
