Different results generated by using the same parameters

[sulicon]

I'm new to 2nd-generation sequencing data analysis. I have found that every time I run newbler, the de novo assemlber from Roche, I get different results... The parameters used are identical in these runs. Is this normal?

update:
It seems that the difference is caused by multiple CPUs used for computation. I had used "-cpu 10" before. If only one CPU used, the result is reproducible, although it's also different from results generated by multiple CPUs...

[flxlex]

What you find is correct, only with -cpu 1 will you get the same results each time. Nice to be able to answer one's own question, eh?

[Torst]

I'm new to 2nd-generation sequencing data analysis. I have found that every time I run newbler, the de novo assemlber from Roche, I get different results... The parameters used are identical in these runs. Is this normal?

There seems to be two ways to get different results from gsAssembler:

1. As you found, the -cpu option with more than 1 cpu gives different results! The only reliable way is with -cpu 1.

2. The order you supply your .SFF files on the command line changes the results, even with -cpu 1 enabled. It seems you need to put your paired reads AFTER your single-end reads.

I am appalled that just important software can be so non-deterministic in its operation!