Most stringent settings for bwa mem

Question

[nanana]

I am trying to optimise bwa mem in order to find the most stringent alignment settings - prioritising accuracy over speed. (See related post).

I am trying to find a set of parameters that would:

• Set the probability of a mismatch/substitution/insertion being detected as equal
• Highly penalise the extension of gaps (insertions and deletions)
• Optimise for accuracy over speed

Here is what I'm currently using:

bwa mem \
    -r1 \  # re-seeding trigger - default [1.5]
    -B6 \  # Mismatch penalty - default [4]
    -O6 \  # Gap open penalty - default [6]
    -E10 \ # Gap extension penalty - default [1]
    -H \   # Hard clipping - default [F]
    ${genome} \
    ${reads}

Does anyone have any suggestions for how I can optimise this to best fit my priorities?

[GenoMax]

Since the settings may be somewhat dependent on data at hand you may have to experiment yourself. Read length, size of database, low diversity are all going to affect final result.

Answer 1

[Istvan Albert]

I recommend avoid using the word "stringent" - there is nothing stringent about the alignment parameters that you set.

It is a set of tuning parameters, it is not any more "stringent" or "accurate" than say a default (usually default parameters are set to attempt to emulate a broad range of biologically relevant situations).

It may very well be that default parameters are not a good fit for you problem, but in both of your posts using words such as "stringent" and "accurate" does nothing more than misleads the reader (and yourself). There is nothing "stringent" about the parameters above.

Just say I want to avoid gap extensions and I prefer multiple opening new gaps to extending existing gaps. Which to be honest is biologically most counterintuitive and if you are using this to model biological phenomena make sure you have the justification for it as it is not how most biology operates.