For the SNP dataset distances ranged from 0.003 and 0.45, with an average of 0.28.
The significance of dataset distances was estimated by constructing a parameter reference distribution by shuffling tissue-specific genes among datasets.
Suitable for small datasets Manhattan distance (Dleft( {i,j} right) = sumnolimits_{i = 1}^{n} {left| {Xi Yi} right|}) Distance based on an absolute value.
Suitable for numeric datasets Jaccard distance (J U,A) = frac{{left| {U cap A} right|}}{{left| {U cup A} right|}}) Used as a measure of similarity.
All the models that used the datasets without distance variables performed quite poorly (Table 1).
We show that the filtering step significantly improves the retrieval performance with most datasets and distance measures.
For all eight datasets, genetic distance was calculated between each pair of samples using the identity by descent (IBS) method implemented in TASSEL.
For both salinity measures, this result remained consistent after correcting for oceanographic connectivity (60 loci and Her14 datasets), geographical distance (though only for Her14 locus), and temperature.
Therefore, we combined four public microarray datasets with Distance Weighted Discrimination [ 14], providing 855 tumors with documented first site of relapse (Supplemental Table 2) [ 15– 15].
Even though the presented analysis exemplifies the proposed method only with a few species, datasets, and distance measures, it can be applied to evaluate a wide variety of data resources and the impact of various parameter settings.
