Procedure to calculate the strength of evidence based on integrated trials results

Results from several trials need to be integrated to make a decision. For example, one gene deletion strain may have been grown with several metabolite nutrients from the same pathway. We then need to answer the general question: "Was this gene responsible for this enzyme or not?". Each trial has an accuracy figure, but how does the combination of these accuracies compare to what is expected by chance (we are concerned about multiple hypothesis testing here).

For each set of trials that we need to integrate:

• Collect the 100 MC sampling results belonging to each trial, as calculated in the "Significance" section of the procedure to automatically interpret trial results for novel biology.
• Repeat 1000 times
• Sample one of the MC results from each trial
• Sum the samples
• Record this sum
• Sort the 1000 sums
• Compare the sum of the real trial accuracies with the sorted list and locate the real sum within this list. Report the probability.

The code that does this can be found in discrimination/integrateTrialResults.py.