[3dem] Reliable cryo-EM Resolution Estimation with Modified FSC

Tue Sep 15 11:03:38 PDT 2020

Dear all,
with this respect I would also like to point the interested readers to the approach of thresholding FSC curves by statistical permutation testing. The thresholding criterion, which was developed by myself and Carsten Sachse, has been termed FDR-FSC and has just appeared in JSB two months ago, here the link https://urldefense.com/v3/__https://www.sciencedirect.com/science/article/pii/S1047847720301520__;!!Mih3wA!SfZSATQc1BkxVdqiXjtIh7xootWI5tUF9u3teN8G3_E1fZYszKTuik5sZWOhTr2MXA$  (https://urldefense.com/v3/__https://www.sciencedirect.com/science/article/pii/S1047847720301520__;!!Mih3wA!SfZSATQc1BkxVdqiXjtIh7xootWI5tUF9u3teN8G3_E1fZYszKTuik5sZWOhTr2MXA$ ) .
In a roughly related direction to Pawel, proper statistical considerations based on construction of confidence intervals and additional considerations of the effects of masking and symmetry on the number of degrees of freedom (or on effective sample sizes as Marin would probably describe it) allow improved resolution estimation. Our algorithm is completely free of any user defined masking (and thus avoid all the drawbacks coming with it) and only needs the respective symmetry that was applied during reconstruction. The noise distribution is then estimated by non-parametric permutation sampling from the data itself.
Our algorithm is available at https://urldefense.com/v3/__https://github.com/MaximilianBeckers/SPOC__;!!Mih3wA!SfZSATQc1BkxVdqiXjtIh7xootWI5tUF9u3teN8G3_E1fZYszKTuik5sZWM1iPs_dQ$  (https://urldefense.com/v3/__https://github.com/MaximilianBeckers/SPOC__;!!Mih3wA!SfZSATQc1BkxVdqiXjtIh7xootWI5tUF9u3teN8G3_E1fZYszKTuik5sZWM1iPs_dQ$ )
Best,
Max 
-------Maximilian BeckersPostdoctoral Researcher Novartis Institutes for Biomedical ResearchNovartis PharmaBasel, Switzerland
Am Di., Sept. 15, 2020 16:49 schrieb "Penczek, Pawel A" :  
 Dear All,
 The currently prevalent single-particle resolution assessment methodology is based on Fourier Shell Correlation (FSC).  However, its use remains controversial as the outcome is subjective, mainly due to the necessity to use smooth-edged masks, which introduce artifacts and make it difficult to support the outcome by statistical analysis.
 In the article Reliable cryo-EM Resolution Estimation with Modified FSC (mFSC) that just appeared on IUCrJ site (https://urldefense.com/v3/__https://doi.org/10.1107/S2052252520011574__;!!Mih3wA!SfZSATQc1BkxVdqiXjtIh7xootWI5tUF9u3teN8G3_E1fZYszKTuik5sZWNErJleAQ$  (https://urldefense.proofpoint.com/v2/url?u=https-3A__doi.org_10.1107_S2052252520011574&d=DwMDaQ&c=bKRySV-ouEg_AT-w2QWsTdd9X__KYh9Eq2fdmQDVZgw&r=yEYHb4SF2vvMq3W-iluu41LlHcFadz4Ekzr3_bT4-qI&m=8P5zavyLYwIWiwPar72EogSpE8PJtfj-B0muENTEnAI&s=Ntn8H-2368avF171GWdLgm2KBTbOhp7T_0jw6Yj8P-U&e=)) I addressed these deficiencies and modified the way FSC is computed.  As a result, the proposed mFSC is free from mask-induced artifacts and thus only binary masks outlining region of interest are used.  Moreover, this modification makes it possible to compute the number of degrees of freedom associated with mFSC coefficients and thus assess the resolution using one-sided Confidence Interval (CI). The use of statistical test prevents users from accepting unreliable results that may stem from using excessively small regions of interest, as in local resolution analysis.  The mFSC program is available at https://urldefense.com/v3/__http://sparx-em.org/sparxwiki/sxresolution__;!!Mih3wA!SfZSATQc1BkxVdqiXjtIh7xootWI5tUF9u3teN8G3_E1fZYszKTuik5sZWMWwcS4fg$  (https://urldefense.com/v3/__http://sparx-em.org/sparxwiki/sxresolution__;!!Mih3wA!U-Ahz68KwuVPUNP6BMZ1CSsJVoTLABdO9IdUnCsBWHrnzowqgjv-nocb8RZ6sdpyyg%24).
 The mFSC is particularly effective in addressing a number of problems:
	* Since there are no mask-induced artifacts, it is possible to assess resolution by entirely eliminating solvent regions, i.e., to use masks that coincide with the structure itself
	* I show that local resolution analysis is of limited value as the spatial resolvability of the outcome is poor. Instead, I suggest using mFSC to compute resolution of biologically relevant segments of the structure or even its secondary elements, such as helices.
	* mFSC makes it possible to compute artifact-free resolution curves for helical structures (for which standard FSC in most cases fails due to reciprocal space artifacts)
	* The Confidence Interval-based mFSC all but eliminates subjectiveness of resolution assessment.
 Regards,
 Pawel A. Penczek, Ph.D.
 Professor
 Department of Biochemistry and Molecular Biology
 The University of Texas - McGovern Medical School
 MSB 6.220
 6431 Fannin
 Houston, TX 77030
 USA

 phone: 713-500-5416
 fax: 713-500-6297
 pawel.a.penczek at uth.tmc.edu (mailto:pawel.a.penczek at uth.tmc.edu)
https://urldefense.com/v3/__http://www.uth.tmc.edu/bmb/faculty/pawel-penczek.html__;!!Mih3wA!SfZSATQc1BkxVdqiXjtIh7xootWI5tUF9u3teN8G3_E1fZYszKTuik5sZWNEE8D0gw$  (https://urldefense.com/v3/__http://www.uth.tmc.edu/bmb/faculty/pawel-penczek.html__;!!Mih3wA!U-Ahz68KwuVPUNP6BMZ1CSsJVoTLABdO9IdUnCsBWHrnzowqgjv-nocb8RZcF7IYlA%24)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20200915/6f99adc3/attachment.html>