We compared the performance of DrugFEATURE to MAPPOD, F-pocket, SiteMap, and DLID using the drug discovery data set, which is derived from Cheng et al.'s original data set.
Using the TCGA discovery data, we identified all genes that were eligible to be included in either the DM or predictable set when |rP| is 0.60 and redundancy is 1; i.e., all genes that are correlated with rP > = 0.60 or rP < = −0.60 to at least 1 other gene.
The validation set was tested using the model constructed on the discovery data, in order to assess model overfitting, and models that were statistically significant in both the discovery and the validation set would be expected to have strong predictive performance.
The best model generated from the discovery data using several of these peaks gave 100% sensitivity, 93.8% specificity, 92.9% PPV and 100% NPV when tested on the independent validation dataset, showing the potential of the test for accurate diagnosis of BTC.
Briefly, we took all SNPs with P < 10−4 in the discovery data and we used these to define association regions (with at least one such SNP) and compiled a list of all genes in these regions.
Statistical analysis was performed using the Visual data discovery software JMP9 (SAS, Cary, NC).
The estimated effect sizes for bipolar disorder-defined score alleles in our ADHD v. control analysis are smaller than that observed when we reanalyse our data using the ISC schizophrenia data-set as discovery data (see the online supplement).
Using the lipidomics data from the discovery series, we derived a model to determine the liver-fat content by applying the artificial-life algorithm and ridge regression as for the NAFLD model.
If researchers make a commercially important discovery using the Celera data, they do not have to give Celera any rights to their inventions.
