In order to repeat our analysis using random sets of 22 residues, we used the set of sequence alignments that include only residues with no gaps as a pool, excluding the 22 SDRs themselves.
In this validation, the methods are examined using random sets of species with dimension 50, 100, 200, and 300, respectively.
This success rate was achieved using random sets of markers chosen only for six-plex PCR because of non-overlapping allele sizes.
Likewise, using random sets of 8 miRNA promoters, we observed that the 8 miRNAs that are over-expressed in monocytes have a factor of 1.58 enrichment of TFBS motifs (p = 0.046; Additional file 5, Figure S3B).
For an anonymous trait where phenotypic accuracy was 0.58, shrinkage increased the average GEBV accuracy from 0.56 to 0.62 (SE < 0.00) when using random sets of 384 markers from a 60K array.
When phenotypic accuracy was above 0.6, it surpassed GEBV accuracy using random sets of 384 SNPs without shrinkage, and the crossover with shrinkage occurred at phenotypic accuracy equal to 0.8.
We then repeated the mapping procedure, using these random sets of positions instead of the SDRs, and reproduced figure 1 for each random set of residues (Figure S1).
Thus, to address this issue and to ensure the validity and stability of the findings, we conducted the analyses using 5,000 random sets of start values and replicated each solution to ensure model stability.
We trained the system using several random sets of known sequences, demonstrating average sensitivity and specificity values of ∼85%.
The three curves correspond to (1) using all 44K markers, (2) using a random set of 384 SNPs with shrinkage, and (3) using 384 SNPs without shrinkage.
The number of genes in each random gene set corresponds to the sizes of the actual gene sets; however, to reduce computational burden, we use random gene sets of pre-specified sizes (1, 5, 10, 25, 50, 75, 100, 125, 150, 200, 250, 300, 400, 500+) that correspond to the (rounded up) sizes of the actual gene sets.
