Our organization (Compendia Bioscience) is working to upgrade from Pipeline Pilot 7.5 to 8.5. During testing, we noticed that one of our protocols produces different results due to rounding.

We have created a simplified test case to demonstrate this behavior. This test case including its required input file is attached as a zip file to this post.

Steps to follow:

1. If you unpack the zip and import RoundingDiscrepancies.xml into both Pipeline Pilot 7.5 and 8.5.

2. Place the input folder in the \$(UserDir) directory for both versions of Pipeline Pilot. This folder contains the input file ndata.txt.

3. After running the protocol a file named quantile-vector.txt will be generated at the base of \$(UserDir).

4. Compare the file quantile-vector.txt that was generated from both versions of Pipeline Pilot. You can also see the results that we experienced in the 7.5_results and 8.5_results directories. The differences in the values that we see between these files is very small (0.986081296673408 vs 0.986081296673407). However, they are used for downstream protocols, and the differences are compounded to the point of significantly altering results.

We would like to understand the cause of the discrepancies between Pipeline Pilot 7.5 and 8.5. Has anyone else experienced rounding discrepancies between the two versions?

Thanks!