We compare the results of the various permutations of input and output variables along with statistical approaches, and discuss the implications of our results in the Discussion section.
The arithmetic difference between input lanes was calculated and the distribution of these differences was determined for all possible permutations of input-input differences, and the mean of these distributions calculated.
Calculation was performed by estimating the sampling distribution using permutation of inputs (cf. permutation tests).
The approach used incorporates the uncertainty inherent in key model parameters by running the model over a large number (almost 4 billion) of permutations of model inputs.
The numeric values are false-discovery rates estimated by 1,000 permutations of the input data.
To account for this lack of exchangeability, Li and Stephens (2003) suggested averaging the PAC likelihood over several (say, between 10 and 20) random permutations of the input haplotypes.
Additionally, the table provides P-value scoreficancestimatesstimates for the enrichment scores, obtained in a non-parametric fashion using 1000 random permutations of the input rankings.
We applied each estimation approach to predict each of the three output measures of life expectancy (i.e., total, at age 60 for males, and at age 60 for females) as a function of various permutations of the input variables described above.
To classify genes as male- or female-biased in a statistically stringent manner, we performed a Bayesian analysis on random permutations of the input data and applied a non-parametric statistical method to classify 467 genes as male-biased and 551 genes as female-biased, which were used for subsequent analysis (Additional file 1: Table S1).
Simulation results together with permutation tests of input gene network structures indicate that the prediction accuracy and robustness property of the two predicted networks using the top-down approach are better than those of the corresponding fully connected networks.
