• sep1 - data is 0.001 subsample; collectContigs's input excludes alignments with quality < 20; also includes oases and trinity
• sep5-0.001 - data is 0.001 subsample; collectContigs's input excludes alignments with quality < 10; also includes oases and trinity
• sep6-0.001 - data is 0.001 subsample; collectContigs's input excludes alignments with quality == 0
• sep7-0.001-sim - data is 0.001 simulation; collectContigs's input excludes alignments with quality == 0; collectContigs was run with our usual input after reconstructing metadata
• sep7-0.001-sim-original.as.input - data is 0.001 simulation; collectContigs was run with the output of rsem-simulate-reads directly (only collectContigs, single-end data)
• sep5-0.01 - data is 0.01 subsample; collectContigs's input excludes alignments with quality == 0; also includes oases and trinity

• oct2.single-full - data is the full Mouse dataset, only single-end data, run with Trinity, collectContigs, and Oases.
• oct2.single-full-sim - data is simulated based on RSEM run on the whole dataset, including all isoforms, and of similar size to the original. Only includes collectContigs output.
• oct2.single-full-samp - Same as -sim, except that when making the collectContigs input, we sample alignments instead of taking the best one.

For all of the above:

• The summary file (or files, if paired-end data was also tried) is available as a csv file in the directory.
• Plots are available.
• Correlation of each column with rsem loglikelihood is recorded in correlations.txt.
• A table showing which assembly achieved the maximum for each column is recorded in maximizers.txt.
• A description of each column is available here.