Both matrices are filtered to leave only homologs between the two organisms and give matrices G 1 and H 1. The cross-coexpression matrix I is given as the mean: I = (G 1 + H 1 ) ∕ 2 This analysis can easily be extended to more datasets by determining a core set of homologs and taking the mean of all individual coexpression matrices.
Data and design matrix were filtered with a unidirectional 1st order Butterworth high pass filter with cut off frequency of 0.05 Hz, according to current recommendations (Bach et al., 2013).
In the BM3D algorithm, similar patches form the 3D matrix and this matrix is filtered in the transform domain with appropriately selected threshold [16].
First the intensity matrix was filtered by dropping away peaks which were detected in less than 5 samples.
Each similarity matrix was filtered to keep the top-n highest similarities per document and then clustered using a combination of graph layout and average-link clustering.
The expression matrix was filtered to contain only those probe sets that can be assigned to a gene name.
Each underlying covariance matrix was filtered, so that only SNPs present and passing QC for the T1DGC study dataset were present.
The methylation data matrix was filtered for all probes with a SNP-annotation or a missing value in any tumor sample (final N = 322,425 probes).
The final genotype matrix was filtered to contain only samples from the domesticated apple, Malus domestica, and ≤20% missing data per SNP and per sample.
This very large result matrix was filtered to retain only those pairs that represented a comparison across platforms (or between a phenotypic variable and a platform) and whose Pearson correlation or MIC values exceeded a threshold of 0.75.
