Mercator's sample projection function allows users to generate similarity scores for each sample in Mercator. A specific .tsv file format must be used to ensure proper processing:
sample1 sample2
"ENSG00000000003.14" 230690 92245
"ENSG00000000005.5" 7694 101
"ENSG00000000419.12" 149758 59561
"ENSG00000000457.13" 45392 42325
"ENSG00000000460.16" 114602 24734
"ENSG00000000938.12" 1611 0
"ENSG00000000971.15" 0 202
"ENSG00000001036.13" 191342 82201
...
In this case we have a two-sample file, but you can include up to five samples per upload. Gene counts should be unnormalized, and we recommend using the script and modified .bed file available at the project Github to generate the counts for optimal performance with Mercator. The script requires binaries for the UCSC tools wigToBigWig and bigWigAverageOverBed, and is a fairly straightforward gene counting protocol similar to the one used in Recount2. The input files must be wiggle files, we recommend using those generated from Rails-RNA or STAR using the --outWigType wiggle and --outWigNorm None options.
cd project_dir
chmod u+x mercator_prep_star.sh
./mercator_prep_star.sh -o output_file.tsv input_file1.wig input_file2.wig input_file3.wig
The output file can be submitted directly to Mercator, although once again only five samples are allowed at a time.
(The links to the UCSC tools are for Linux, here are the MacOS binaries for wigToBigWig and bigWigAverageOverBed)