Corrects and transforms two-colour microarray differential expressions (D.B. Baird).
|What to print (
||How to correct for spot foreground for background values (
||Type of transformation to apply to the red/green ratios (
||Minimum value per channel; if
||Use a single minimum value per spot rather than per slide (
||Constant to add to red and green foreground values; default 0|
||Degrees of freedom to use for loess smoothing of background; default 20|
||Red foreground values per spot|
||Green foreground values per spot|
||Red background values per spot|
||Green background values per spot|
||Standard deviation of red background|
||Standard deviation of green background|
||Defines the slide to which each spot belongs for smoothing, or per slide minima|
||Defines the row position of each spot for background smoothing|
||Defines the column position of each spot for background smoothing|
||Saves the differential expression per spot|
||Saves the intensity of each spot|
||Saves the corrected red values per spot|
||Saves the corrected green values per spot|
MACALCULATE calculates log-ratios from a two channel microarray. There are options for background correction and to set the channels to a minimum value to avoid large variation when the values in either channels get close to zero. The ratios are logged as the variation usually increases with the size of the mean. However, although the variation may be more uniform with the logged ratios, it may still show considerable heterogeneity with different levels of intensity.
GFOREGROUND parameters supply the values of the red/Cy5 and green/Cy3 foreground for each spot. (Cy5 and Cy3 are the technical names for the red and green fluorescent dyes.) These may be single variates. The
SLIDES parameter then supplies a factor indicating the slide from which each spot was read. Alternatively, they may be pointers, containing a variate for each slide. The
SLIDES parameter can then be omitted, or it can supply a text giving a label for each slide. The
GSDBACKGROUND parameters must be in the same format as the
||subtract the red and green backgrounds, which must then be supplied by the
||perform a two-dimensional loess smoothing of the background values, and uses these for the correction: the
BMETHOD is unset, the red and green backgrounds are subtracted if available. The
DF option specifies the number of degrees of freedom for the smoothing; if this is unset, the default is to use the square root of the number of spots observer per slide. Note that smoothing can be time consuming for large slides.
Log-ratios are undefined where spots may have foreground levels that are below their background levels, as you cannot take the log of a negative number. If this happens for both red and green, then there is no valid information on the level of differential expression and the log-ratio must be set to a missing value. However, there may be some useful information when one channel is above background and the other below and, if these log-ratios too are set to missing values, probes with a high level of differential expression may be missed. Some image analysis packages set a very large constant log-ratio for this, but this can bias the results as no differentiation is made between cases where the other colour is just above background, and other where the other colour is significantly above background.
You can set the
MINIMUM option to a positive value to apply minimum values to the colours (for the default value of zero, no minimum values are applied). In the simplest situation, any red or green value less than
MINIMUM is set to
MINIMUM. For the alternatives, you need to supply variates giving the standard deviations of the red and green backgrounds around each spot, using the
GSDBACKGROUND parameters. Then, if option
PERSPOTMINIMUM=yes, the background standard deviation for each colour is multiplied by
MINIMUM, so that the minimum value depends on the quality of the background around each spot. With the default,
MINIMUM is multiplied by the background standard deviation for each colour, averaged within each slide.
CONSTANTVALUE option can be used to reduce variability at the low end of the intensity range, as often the ratios become unstable as the foreground gets close to the background level. Adding a constant to both red and green will stabilize the log-ratios at low intensities, but with a potential loss of sensitivity for detecting differential expression.
CONSTANTVALUE would normally be a positive value, although a negative value could be used if it were thought that the image analysis package was underestimating the background values. By default
LOGRATIOS parameter can save the calculated log-ratio or generalized log-ratio transformation (as requested by the
TRANSFORMATION option) of the red and green values for each spot. The
INTENSITIES parameter can save the intensity of each spot:
Intensity = (Log(Red) + Log(Green)) / (2 × Log(2))
GCORRECTED parameters save the corrected red and green values (i.e. after the application of any background corrections, minimum values and constants) for each spot. If they have already been defined, the data structures supplied by
GCORRECTED must have the same type as that specified by the
RFOREGROUND parameter (i.e. variates if
RFOREGROUND is a variate, and pointers if
RFOREGROUND is a pointer).
Commands for: Microarray data.
CAPTION 'MACALCULATE example'; STYLE=meta ENQUIRE CHANNEL=-1; EXIST=check; NAME=\ '%GENDIR%/Data/Microarrays/Data13-6-9.gwb' IF check SPLOAD '%GENDIR%/Data/Microarrays/Data13-6-9.gwb' CAPTION 'MACALCULATE: no background correction'; STYLE=minor MACALCULATE [PRINT=summary; TRANSFORM=log; BMETHOD=none]\ RFOREGROUND=F1_Mean; GFOREGROUND=F2_Mean; SLIDES=Slide;\ LOGRATIO=logRatio; INTENSITIES=Intensity CAPTION 'MACALCULATE with background correction'; STYLE=minor MACALCULATE [PRINT=summary; TRANSFORM=log; BMETHOD=subtract;\ MINIMUM=0.5; PERSPOTMIN=No]\ RFOREGROUND=F1_Mean; GFOREGROUND=F2_Mean; SLIDES=Slide;\ RBACKGROUND=B1_Median; GBACKGROUND=B2_Median;\ LOGRATIO=logRatio; INTENSITIES=Intensity;\ RSDBACKGROUND=B1_SD; GSDBACKGROUND=B2_SD ELSE CAPTION 'Microarray example datasets have not been installed.' ENDIF