[3dem] [ccpem] to resolution or not to resolution

[Marin van Heel]

Dear Teige,

You state  "... whether you use a 1/2 bit FSC criterion  ... or a 0.143 
FSC criterion justified by Peter, Tony and Richard... " as it there is a 
equal choice.
On the contrary:  we show the "0.143 FSC criterion" to be WRONG , 
because it is based on flawed ("sloppy") statistics! Unfortunately this 
applies to almost all resolution criteria in use in cryo-EM (see our 
BioRxiv paper). In science you have to simply accept you made a mistake, 
lick you wounds, and move on.  The discussion during the last two weeks 
was primarily about the consequences of under-sampling the data  a 
closely related but equally widely misunderstood topic.

Chiara, the FSC is a 1D curve and its behavior below the 1/2 bit 
threshold may be an important thing to look at 
(http://journals.iucr.org/m/issues/2017/05/00/kf5002/index.html).

Have fun,

Marin

PS:  this slide is from my Wiley lecture in April 2017 and which I also 
used in  my presentation in Stockholm on the 9th of Dec 2017 at the 
cryo-EM symposium during the Nobel 2017 festivities.





On 08/09/2018 07:26, Teige Matthews-Palmer wrote:
> Dear Chiara,
>
> My (non-expert) take on it is that, whether you use a 1/2 bit FSC 
> criterion (asymptotic to 0.1716), advocated by Marin: 
> https://www.biorxiv.org/content/biorxiv/early/2017/11/24/224402.full.pdf or 
> a 0.143 FSC criterion justified by Peter, Tony and Richard in this 
> appendix: 
> https://www.sciencedirect.com/science/article/pii/S0022283603010222?via%3Dihub#APP1 
> <https://www.sciencedirect.com/science/article/pii/S0022283603010222?via=ihub#APP1> , 
> these fall-off thresholds require some assumptions to be met. E.g. FSC 
> is between two independent half-maps; FSC vs resolution curve is a 
> ‘nice shape’  - a threshold shouldn’t replace looking at the curve, 
> which should be high before a single fall-off without big oscillations.
> There is a debate going on about whether reporting a resolution beyond 
> particular fractions of Nyquist is valid, but your negative stain 
> reconstruction will be too far from Nyquist for you to worry about that.
> Whereas a half-map FSC for high-res cryo is trying to determine a 
> threshold where correlated signal can be interpreted as scattering by 
> the molecules, negative stain is giving you a lot of signal from 
> scattering by variable stain envelopes.
> I think so long as the FSC curve is not a weird shape or oscillating, 
> you should be able to use the same FSC criterion as cryo, for low-res. 
> I’d love to hear more experienced views on this.
>
> Another quality measure could be to look at your map’s anisotropy. 
> Relion for example automatically produces a 3D histogram of assigned 
> particle angles, and Dmitry Lyumkis has shared his group’s program for 
> measuring anisotropy with FSC: 
> https://mail.ncmir.ucsd.edu/pipermail/3dem/2017-September/005210.html
>
> Finally, you might want to know if the shape of your 3D map at low-res 
> is correct, which is definitely not a given. You can collect 
> tilt-pairs (or do tomography): 
> https://www.sciencedirect.com/science/article/pii/S0959440X15000925 but 
> I think there have been other validation methods suggested that don’t 
> require extra data, but I can’t find the paper. T_T Does anyone know?
>
> All the best,
> Teige
>
>
>> On 8 Sep 2018, at 01:00, Dmitry Lyumkis <dlyumkis at SALK.EDU 
>> <mailto:dlyumkis at SALK.EDU>> wrote:
>>
>> Weekend FSC spar 2.0?
>>
>>
>> On Sep 7, 2018, at 11:34 AM, Chiara Rapisarda 
>> <c.rapisarda at IECB.U-BORDEAUX.FR 
>> <mailto:c.rapisarda at IECB.U-BORDEAUX.FR>> wrote:
>>
>>> Dear community members,
>>>
>>> I would like to know what is the state of the art for reporting 
>>> negative stain reconstruction resolution.
>>> Is it best to not include it or to use the 0.5 FSC cut off. I 
>>> definitely don't want to use the 0.143 cut-off, but I thought that 
>>> it is important to give an idea of the quality of the reconstruciton 
>>> by using some form of quantitative measure.
>>>
>>> Is there an agreement on what to do?
>>> Thank you for any feedback I will receive.
>>>
>>> Chiara Rapisarda
>>>
>>> Post doc
>>> Université de Bordeaux, CNRS, INSERM
>>> c.rapisarda at iecb.u-bordeaux.fr <mailto:c.rapisarda at iecb.u-bordeaux.fr>
>>> Tel. +33 (5) 4000 3617
>>>
>>> Institut européen de chimie et biologie
>>> 2, rue Robert Escarpit
>>> 33607 Pessac, France
>>> www.iecb.u-bordeaux.fr <http://www.iecb.u-bordeaux.fr/>
>>>
>>>

-- 
==============================================================

     Prof Dr Ir Marin van Heel

     Laboratório Nacional de Nanotecnologia - LNNano
     CNPEM/LNNano, Campinas, Brazil
     email:  marin.vanheel(A_T)gmail.com
             marin.vanheel(A_T)lnnano.cnpem.br
     and:    mvh.office(A_T)gmail.com

--------------------------------------------------

I receive many emails per day and, although I try,
there is no guarantee that I will actually read each incoming email.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20180908/96bc7545/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mpfmbbkmoikkogjc.png
Type: image/png
Size: 1377564 bytes
Desc: not available
URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20180908/96bc7545/attachment-0001.png>