Open animesh opened 1 year ago
Hi Ani,
Reproducibility is guaranteed if you use the same commands. Using library creation on the fly is not supposed to yield identical results as searching with .predicted.speclib, this is for technical reasons. In general, searching with .predicted.speclib is the recommended way.
Best, Vadim
Since --reanalyse
creates a library and re-searches using that library, how can we recreate the second pass search using the output library from the first pass? E.g., processing a set of files with --reanalyse
creates both report-first-pass.tsv
and report.tsv
(the second pass) and a report-lib.tsv
.
If I take the report-lib.tsv
and search the same set of files without --reanalyse
, I am unable to reproduce the report.tsv
. Is there a way to reproduce the second pass? Any suggested flag changes? I assume DIA-NN implicitly does a tighter search in the second flag and we need to some how enable that search from the command line?
I am trying to compare results between library-free search report.pg_matrix.tsv.txt
and a search using the library generated from previous search report.pg_matrix.tsv.txt
and despite couple of IDs missing from each other compare.txt looks like spearman/rank-correlation between results is about ~99.9%
i am wondering about the differences, specifically the marked P11142 , also if there a way to reduce this randomness to make result fully reproducible between two searches?