How to understand different results of AmpliconArchitect in different servers?

[JFanbio]

Hi,

I Ran the AS-pipeline for the bam that my colleague has run by AA(1.2_r2) and AC (0.4.13) but got different results, then I considered it as a version problem.
So, I install the AA and AC as the same version as my colleague. I also use the same ref, the same CNV seeds file, and the same Mosek license. However, I still failed to get the same results as my colleague.
Comparing the results, it seems my results were a subset of my colleague's.
I'd like to know what's the reason for the difference, and if there is anything I need to check more carefully.
Thank you!

Best,
Jian

[jluebeck]

Hi Jian,

Thank you for reaching out about this. I need a bit more information to try and debug this.

For instance, how can I reproduce the outputs for the sample in question? Is there a way you can share the bam file in question?

Alternatively, can you send me the complete outputs (not bam) from your run and your colleague's run both for the latest versions of the tool?

Thanks,
Jens

[JFanbio]

Hi Jens,
My colleague didn't run the latest version, since she worried that some unknown factor would influence her results. What I can tell you is I copied the tool sources from her because AA(1.2_r2) and AC (0.4.13) can't be downloaded online anymore.
I can send you my results. For the results of her, I need to talk with her.
Thank you!

Best,
Jian

[JFanbio]

Hi Jens,

I have run AA and AC for a bam twice, with totally the same parameters, but got different results, like a gene was in ecDNA for first run but in an unknown part for a second run.
How did this happen, can you help me with that?

Best,
Jian

[jluebeck]

Hi Jian,

The older version of the tools uses a non-deteministic random seed for downsampling. Since 1.3 and on it has been deterministic. That could theoretically cause the issue. I'm not sure if you have already tried to send me output files to examine, but you can email me at jluebeck [at] ucsd.edu. It is a bit hard for me to know what is going on without seeing your outputs or logs.

I will also point out that AA 1.2.r2 and AC 0.4.13 are fairly old versions of the tools. I recommend you upgrade to the latest. I have never encountered the issue you described in the later versions of the tools.

Thanks,
Jens

[JFanbio]

Hi Jens,

I think I can make sure the downsampling is the reason. Thank you!
I suggest the default value of this parameter set as -1 will be better, since the higher depth sequencing data helps to get more complex rearrangement events, it's more important than speeding up.
Always thank you for your patient reply!

Best,
Jian