aggregateLoadings.RdFirst, this function identifies which loadings are within the region set. Then the loadings are used to score the region set according to the `scoringMetric` parameter.
aggregateLoadings(loadingMat, signalCoord, regionSet, PCsToAnnotate = c("PC1", "PC2"), scoringMetric = "regionMean", verbose = FALSE)
| loadingMat | matrix of loadings (the coefficients of the linear combination that defines each PC). One named column for each PC. One row for each original dimension/variable (should be same order as original data/signalCoord). The x$rotation output of prcomp(). |
|---|---|
| signalCoord | a GRanges object or data frame with coordinates for the genomic signal/original data (eg DNA methylation) included in the PCA. Coordinates should be in the same order as the original data and the loadings (each item/row in signalCoord corresponds to a row in loadingMat). If a data.frame, must have chr and start columns. If end is included, start and end should be the same. Start coordinate will be used for calculations. |
| regionSet | A genomic ranges (GRanges) object with regions corresponding to the same biological annotation. Must be from the same reference genome as the coordinates for the actual data/samples (signalCoord). |
| PCsToAnnotate | A character vector with principal components to include. eg c("PC1", "PC2") These should be column names of loadingMat. |
| scoringMetric | A character object with the scoring metric. "regionMean" is a weighted average of the absolute value of the loadings with no normalization (recommended). First loadings are averaged within each region, then all the regions are averaged. With "regionMean" score, be cautious in interpretation for region sets with low number of regions that overlap signalCoord. The "simpleMean" method is just the unweighted average of all absolute loadings that overlap the given region set. Wilcoxon rank sum test ("rankSum") is also supported but is skewed toward ranking large region sets highly and is significantly slower than the "regionMean" method. For the ranksum method, the absolute loadings for loadings that overlap the given region set are taken as a group and all the loadings that do not overlap the region set are taken as the other group. Then p value is then given as the score. It is a one sided test, with the alternative hypothesis that the loadings in the region set will be greater than the loadings not in the region set. |
| verbose | A "logical" object. Whether progress of the function should be shown, one bar indicates the region set is completed. Useful when using aggregateLoadings with 'apply' to do many region sets at a time. |
a data.table with one row and the following columns: one column for each item of PCsToAnnotate with names given by PCsToAnnotate. These columns have scores for the region set for each PC. Other columns: cytosine_coverage which has number of cytosines that overlapped with regionSet (or in the general case, coordinates from signalCoord that overlapped regionSet) region_coverage which has number of regions from regionSet that overlapped any coordinates from signalCoord, total_region_number that has number of regions in regionSet, mean_region_size that has average size in base pairs of regions in regionSet, the average is based on all regions in regionSet and not just ones that overlap.
data("brcaMCoord1") data("brcaLoadings1") data("esr1_chr1") rsScores <- aggregateLoadings(loadingMat=brcaLoadings1, signalCoord=brcaMCoord1, regionSet=esr1_chr1, PCsToAnnotate=c("PC1", "PC2"), scoringMetric="regionMean")