We selected the fragment distribution derived from ChIP-Seq data of 9 histone markers around TSS in 27,000 protein coding genes as positive set and those on 10,000 random positions as negative set.
The histone binding pattern nearby the TSSs of all protein coding genes is retrieved as positive set, and the ones of 10,000 random positions are retrieved as negative set.
A very powerful separation hyperplane can be created using these GC-rich promoters as positive set and the, generally AT-rich, intronic sequences as negative set.
A total of 107 cardiovascular disease genes and 81 endocrine disease genes are collected, respectively (both treated as positive set P).
As positive set we used proteins whose folding kinetics and thermodynamics have been studied in vitro [71 non-redundant entries (Tartaglia and Vendruscolo, 2010)].
It means for developing models for predicting IgA epitopes; we used IgA epitopes as positive set and the rest of the epitopes (IgE, IgG and non-BCE) as negative set.
Because a reference set of known interactions is not available for S. meliloti, here we consider as positive set those functional linkages belonging to the same COGs functional category [ 37, 60].
To identify disease genes, traditional machine learning methods typically build a binary classification model using confirmed disease genes as positive set P and unknown genes as negative set N. The negative set N is noisy because the unknown gene set U contains some unknown disease genes.
Qi et al. used MIPS and TAP06 as the positive sets.
