Running Abyss On Cluster With Time Limitations

Question

[Raygozak]

Hi, i want to do an assembly but the cluster that i intend to use has a limit on the maximum time a given job can run, is it possible to run abyss for certain number of iterations or time and start right where it left the previous time?. If not, does any other assembler do this? . Also, is it valid to divide the reads, do assembly on each subset, and the do assembly on the results of each chunk?

Thanks

Answer 1

[Shaun Jackman]

Farhat is right, you can resume an ABySS assembly where it left off, but there's no need to edit the abyss-pe script. By default, it will resume an aborted assembly where it left off (due to it's being a Makefile script). You can use the --dry-run option of abyss-pe (it's actually an option of make) to see which commands will be run.

Cheers, Shaun

[Farhat]

Ah. Wish I had known this. Would have saved me a bunch of time :).

Answer 2

[Farhat]

It is possible to get abyss to start where it left off (to some extent). When you run abyss, take a look at the output log and see which commands finished successfully. The command which was running while you ran out of time is where you should start off.

Assuming you are doing a paired-end assembly using abyss-pe, then open abyss-pe in an editor and comment out the commands that have successfully finished. The commands start with the comment

# Assemble unitigs

Before that are definitions and so on which should not be touched. Of course, for this to work at least the first command should finish successfully in the time limit that you have.