Critically, we were interested in the conditional probability of targets ranked 2nd given they were not ranked 1st.
The proportions of targets ranked most, next most, and least likely to have been seen before, are reported in Table 1.
All 1923 targets ranked by their PS α scores across all 182 compounds over the ten models derived from the ten different partitions of the ChEMBL dataset.
In order to verify that the encoding manipulation worked, we compared the average hit rate (i.e. targets ranked 1st) for the weak targets to the average hit rate for the strong targets using a dependent t-test.
PPB searches through 4613 groups of at least 10 same target annotated bioactive molecules from ChEMBL and returns a list of predicted targets ranked by consensus voting scheme and p value.
It suggests that the therapeutic targets can be well identified in the top three predictions, and considering more targets ranked outside the top three would result in a significant number of false predictions.
Let's start with the obvious: His 106 catches (on 155 targets) rank second in the league and his 1,284 yards (he didn't play in Week 17) slot him fourth.
Specifically, 4 out of 4 of the top 20 targets, 2 out of 5 targets ranked between 21 and 60 and 0 out of 14 targets ranked between 61 and 1160 showed >2-fold enrichment (Fig. 1A).
Core genes or miRNA targets ranked at the top or bottom of each pathway gene list were most significantly altered between tumors and normal adjacent tissues (called leading edge in GSEA).
