search

ispyb / ispyb-database-modeling

4 stars 3 forks source link

Add ISa to ISPyB strategy statistics #1

Open olofsvensson opened 7 years ago

olofsvensson commented 7 years ago

The following results is part of the XDS CORRECT.LP file:

 ******************************************************************************
    CORRECTION PARAMETERS FOR THE STANDARD ERROR OF REFLECTION INTENSITIES
 ******************************************************************************

 The variance v0(I) of the intensity I obtained from counting statistics is
 replaced by v(I)=a*(v0(I)+b*I^2). The model parameters a, b are chosen to
 minimize the discrepancies between v(I) and the variance estimated from
 sample statistics of symmetry related reflections. This model implicates
 an asymptotic limit ISa=1/SQRT(a*b) for the highest I/Sigma(I) that the
 experimental setup can produce (Diederichs (2010) Acta Cryst D66, 733-740).

     a        b          ISa
 7.512E-01  2.195E-03   24.62

The 'a', 'b' and 'ISa' values should be added as new columns for the ISPyB AutoProcScalingStatistics table. Although the 'ISa' value can easily be calculated given the 'a' and 'b' values we should store also the calculated 'ISa' value.

Other programs (e.g aimless) also calculates ISa values. These values should also be stored in the AutoProcScalingStatistics table.

olofsvensson commented 7 years ago

The following ISa results can be found in XSCALE.LP:

******************************************************************************
    CORRECTION PARAMETERS FOR THE STANDARD ERROR OF REFLECTION INTENSITIES
 ******************************************************************************

 The variance v0(I) of the intensity I obtained from counting statistics is
 replaced by v(I)=a*(v0(I)+b*I^2). The model parameters a, b are chosen to
 minimize the discrepancies between v(I) and the variance estimated from
 sample statistics of symmetry related reflections. This model implicates
 an asymptotic limit ISa=1/SQRT(a*b) for the highest I/Sigma(I) that the
 experimental setup can produce (Diederichs (2010) Acta Cryst D66, 733-740).
 Often the value of ISa is reduced from the initial value ISa0 due to systematic
 errors showing up by comparison with other data sets in the scaling procedure.
 (ISa=ISa0=-1 if v0 is unknown for a data set.)

     a        b          ISa    ISa0   INPUT DATA SET
 1.063E+00  6.797E-03   11.76   12.36 XDS_ASCII.HKL

 FACTOR TO PLACE ALL DATA SETS TO AN APPROXIMATE ABSOLUTE SCALE 0.114318E+04
 (ASSUMING A PROTEIN WITH 50% SOLVENT)
olofsvensson commented 7 years ago

The idea is to add the following columns to the AutoProcScalingStatistics table:

From XDS (CORRECT.LP):

From XSCALE (XSCALE.LP):

The current ISa column in AutoProcScalingStatistics should be renamed to XDS_ISa.

antolinos commented 7 years ago

As nobody has said nothing against. I think we should implement this.