Setting cutoff for blast with low evalue

Greetings! We would like to run blasts with fairly high stringency, e.g. e values <e-50
When we enter values in the “Set expectation value cutoff” field, it seems we are restricted to using strings of zeros. Is there a way we can enter cutoffs in scientific notation.

1 Like

Hi @tom_abrams_lab

The BLAST+ tools available in Galaxy are wrappers around the source tools from NCBI.

Meaning, the parameter inputs are determined by the underlying tool. Filters for evalue are submitted and interpreted in decimal format.

Reference: https://www.ncbi.nlm.nih.gov/books/NBK279684/#appendices_Options_for_the_commandline_a

Hope that helps!

Thanks for the info! Will the blast work properly entering a long string (e.g. 50) zeros for the e value cutoff?

1 Like

@tom_abrams_lab

I honestly don’t know whether BLAST+ itself will accept that value. You could check the NCBI documentation for the tool and their public forums. If it does, a query with that low of a sensitivity threshold might be unlikely to succeed when working at a public Galaxy server. The intermediate results could grow into a very large memory cache and exceed the capabilities of the clusters. Much would depend on the query, the target database, and other parameters of course. You can certainly try.

If you are working at your own Galaxy, and are able to allocate sufficient resources, then any limits the tool can natively compute/handle should be possible using the same tool as wrapped for Galaxy.

So, the best answer I have is to test it out. Maybe on a small test query first, then scale up for the full analysis.

Hum, looking at the wrapper code, it does have a test for an evalue threshold in scientific notation now – as low as 1e-41 for megablast. Low evalues are present in tests for Blastx, and they are in scientific notation. I hadn’t noticed that before!

Maybe just test it out? This tool is maintained by a dedicated developer who is responsive about questions and requests if open issues (bugs, enhancement requests) are reported.

So give it a try with scientific notation and follow the format as the tests use, and perhaps open an issue against the wrapper repository if you run into technical problems. If you get an error that looks to be server-side, or even if you are not sure, feel free to post back here. Include the public server URL where you are working for context. Or post back the stderr and stdout if working at your own server and we can try to help triage what is going wrong.

(I don’t personally use BLAST+ often anymore … but did about ~20 years ago (line-command)! Apologies for not being current on the exact status of the tool/wrapper :slight_smile: I’m just looking up the same resources online as you probably are to answer for this)

Thanks much Jenna. This scientific notation for the cutoff e-value worked perfectly. I am not sure whether there is a limit, but I tried both e-41 and e-60. What was critical is that I needed to include the1 (as in 1e-50).

1 Like