Differences

This shows you the differences between two versions of the page.

--- lunchtime_byte_2023 [2022/12/21 15:10]
maly structure
+++ lunchtime_byte_2023 [2023/01/05 12:39] (current)
maly cc* English
@@ Line 1: / Line 1: @@
-====== PAIREF in CCP4 - lunchtime byte ======
+====== PAIREF in CCP4 - CCP4SW Lunchtime Byte ======
- Jan 2023
+th Jan 2023
+[[https://pairef.fjfi.cvut.cz/docs/pairef_poli_ccp4sw2023/20230105_ccp4sw_pairef.pdf|{{pdf_icon.png?30x35}}]] [[https://pairef.fjfi.cvut.cz/docs/pairef_poli_ccp4sw2023/20230105_ccp4sw_pairef.pdf|Download presentation in PDF]]
 In this short tutorial, we submit a paired refinement job using the //PAIREF// program, now distributed in the [[http://www.ccp4.ac.uk/|CCP4 suite]]. You can use our example data from interferon gamma from //Paralichthys olivaceus// (POLI)  (([[https://www.sciencedirect.com/science/article/pii/S1050464818302651|Zahradnik et al. (2018). Fish Shellfish Immunol. 79:140–152]])), PDB entry [[https://www.rcsb.org/structure/6F1E|6F1E]].
@@ Line 44: / Line 46: @@
 The results should look similar to ours: [[https://pairef.fjfi.cvut.cz/docs/pairef_poli_ccp4sw2023/PAIREF_poli_ccp4sw2023.html]].
-//PAIREF// ran all the calculation and did also an automatic suggestion of an optimal high-resolution cut off. Let's check the table in the top of the HTML log file:
+//PAIREF// ran all the calculation and did also an automatic suggestion of an optimal high-resolution cut off. Let's check the table on top of the HTML log file:
 {{ :poli_ccp4sw2023_verdict_table.png?nolink |}}
-The first thing that should be checked is whether the refinements have converged. Scroll at the very bottom of page, here you can see plots of Rwork and Rfree vs. refinement cycle. We can conclude that all the refinements have converged.
+The suggestion is based on the results plotted in the following graphs. Overall Rfree decreased for all the three high-resolution shells that denotes model improvement:
-Rfree decreased after the addition of shells 2.3-2.10 Å and 2.1-2.0 Å:
+{{ :poli_ccp4sw2023_r-values.png?nolink |}}
-{{ :poli_webinar_r-values.png?nolink |}}
+However, we should take into account multiple criteria. Since a perfect model gives an R-value of 0.42 against random data (i.e. pure noise) – assuming non-tNCS (translational non-crystallographic symmetry) data from a non-twinned crystal ((Evans, P. R. & Murshudov, G. N. (2013). Acta Cryst. D69, 1204-1214.)) – a higher R-value in the (current) high-resolution shell indicates either the involvement of high-resolution data without information content (the data are even worse than noise), or poor quality of the model, or the presence of tNCS. This is indicated for the shell  2.0-1.9 Å.
+{{ :poli_ccp4sw2023_rfree.png?nolink |}}
-Since a perfect model gives an R-value of 0.42 against random data (i.e. pure noise) – assuming non-tNCS (translational non-crystallographic symmetry) data from a non-twinned crystal ((Evans, P. R. & Murshudov, G. N. (2013). Acta Cryst. D69, 1204-1214.)) – a higher R-value in the (current) high-resolution shell indicates either the involvement of high-resolution data without information content (the data are even worse than noise), or poor quality of the model, or the presence of tNCS. This is indicated for the shell  2.0-1.9 Å.
+CC* is a model-independent measure of noise is in the diffraction data. For this data set, CC* is higher than CCwork in the whole resolution range, except the shell 2.0-1.9 Å where CC* is undefined due to negative CC1/2. That means overfitting was not indicated but the shell 2.0-1.9 Å should be discarded because these data are very noisy. Note that to access overfitting, it is not needed to use test set, so the comparison of CC* with CCwork is much better than with CCfree as CCwork is calculated on more data.
-{{ :poli_webinar_rfree.png?nolink |}}
+{{ :poli_ccp4sw2023_ccwork.png?nolink |}}
-CC* a model-independent measure of noise is in the diffraction data. CC* is higher than CCwork in whole resolution range (except the shell 2.0-1.9 Å where CC* is undefined due to negative CC1/2. Thus, the overfitting was not indicated. To access overfitting, it is not needed to test set, so the comparison of CC* with CCwork is much better then with CCfree as CCwork is calculated on more data.
+We should also check whether the refinements have converged. Scroll at the very bottom of page, here you can see plots of Rwork and Rfree vs. refinement cycle. We can conclude that all the refinements have converged, indeed.
-{{ :poli_webinar_ccwork.png?nolink |}}
-**We can conclude that the high-resolution cutoff can be set to 2.0 Å.**
+**We can conclude that the high-resolution limit of the data is ca. 2.1 Å.**
 Merging statistics:
@@ Line 73: / Line 75: @@
       2.10   2.00  49772   4021   12.38  98.51       0.3     0.1   16.989   17.721    4.963   0.027   0.008   0.2293
       2.00   1.90  35920   4046    8.88  81.18       0.1     0.0   41.435   43.993   14.417  -0.132  -0.016   N/A</code>
+==== Stay tuned, CCP4i2 interface for PAIREF is coming soon! ====
+{{ :pairef_ccp4i2_dev.png?nolink |}}
 ==== Run PAIREF in command line ====

User Tools

Differences

Page Tools

PAIREF

Site Tools