Currently, replicate samples are summed before comparison with G-test. This results in larger counts and increased sensitivity. Using the mean instead could reduce FPs from low-count transcripts.
[ ] Benchmark the mean.
[ ] Make it an option, default to whichever seems best in the benchmark.
Currently, replicate samples are summed before comparison with G-test. This results in larger counts and increased sensitivity. Using the mean instead could reduce FPs from low-count transcripts.