Plate Raw Data Transformer: 1536 to 384 transformation errors

[JenniferSplaine]

Sean Where are we with 1536 in the raw data transformer. I thought it was working and we tested last year. Then the raw data transformer got a rewrite. was the 1536 part updated? As luck would have it i have a 1536 screen currently. Tried to use the transformer this evening and its not working like its supposed to. Putting the data in the 4 384 wells plates wrong and each plate has only 369 wells not 384. Can you look into this? I can hand format for now but it will have to be fixed as we have 2 1536 well screen looming

thanks Jen

[hmsiccbl]

A couple of issues were identified:

• a bug where the matrix reader was storing the row labels with the row arrays. Later code offset the counter to skip the labels, but this was incorrect, the labels should not have been stored. This error did not manifest until a deconvolution (non 386 to 386 well) transformation was done.
• Correlation order necessary to deconvolute was not available.

Summary: When deconvoluting (going from 1536 to 384, for instance), the collation ordering always must have "plates-quadrants" last. What this means is that the 4 quadrants of the 1536 plate must first divide into the first 4 (386 well) plates given, the next 4 quadrants to the next 4 386 plates, and so forth.

So matrix 1 deconvolutes into 4 matrices; which are assigned to the first 4 plates.

To support deconvolution, the following collation orderings were added:

replicates->conditions->readouts->plates->quadrants replicates->readouts->conditions->plates->quadrants readouts->replicates->conditions->plates->quadrants readouts->conditions->replicates->plates->quadrants conditions->readouts->replicates->plates->quadrants conditions->replicates->readouts->plates->quadrants