Closed marcelm closed 3 months ago
I am unsure about read length 75 and 100: With these new settings, would decrease when going from read length 75 to 100, which is a bit unexpected.
Let's try with (19, 15, 0, 3).
Once @Itolstoganov PR is ready, let’s integrate these combinations and then I'll set off a larger benchmark.
Here are the results from the parameter optimization script for all read lengths, run on the multi-context seeds branch.
How the script was run
For reproducibility, I plan to make the script publicly available at https://github.com/NBISweden/strobealign-evaluation.
I ran it on strobealign commit c4a7f61 (from PR #388). This was the command:
--mapping-rate-slack 1
means that all parameter combinations are skipped for which the mapping rate of one of the datasets is reduced by more than 1 percentage points.-x
means that accuracy for mapping-only mode is optimized (i.e., running strobealign with-x
). However, when the list of parameters has been found, the script reports also results for extension-alignment mode so that one can check how its accuracies change.Suggested changes
Parameters are given as a tuple $(k, s, l, u)$.
I am unsure about read length 75 and 100: With these new settings, $k$ would decrease when going from read length 75 to 100, which is a bit unexpected.
[^1]: Already added as new canonical read length
Detailed results