Value of 100 in B1_mean in affygw6.hmm

[isobel16]

Hi there,

I am attempting to train my own HMM, initialised with the parameters from affygw6.hmm, as the array used in my data was a custom Axiom array. However, I have noticed that in the B1_mean values:

[19] "B1_mean:"
[20] "-1.704554 -0.526063 0.000000 100.000000 0.342932 0.597818 "

The 100 value seems out of place/biologically implausible. Please could you advise on this?

Also, I wondered if you could clarify why there are 6 values for B1 and 5 for B2. Thanks! If you had any general advice on optimising this hmm file, as I see the --train option in the detect_cnv.pl script is not recommended for use, it would be greatly appreciated.

[kaichop]

The 100 value is used there to disable the state 4 (which is copy neutral LOH), since many users do not care about it. If you need it, you can manually change to 0 and the prediction will be in output file.

…

On Thu, Dec 5, 2024 at 7:35 AM isobel16 ***@***.***> wrote: Hi there, I am attempting to train my own HMM, initialised with the parameters from affygw6.hmm, as the array used in my data was a slight variation of this array. However, I have noticed that in the B1_mean values: [19] "B1_mean:" [20] "-1.704554 -0.526063 0.000000 100.000000 0.342932 0.597818 " The 100 value seems out of place/biologically implausible. Please could you advise on this? Thanks! — Reply to this email directly, view it on GitHub <#122>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABNG3OEXH77H43HUKR5PPUD2EBCAVAVCNFSM6AAAAABTCLUV2KVHI2DSMVQWIX3LMV43ASLTON2WKOZSG4ZDAMRZHEYTKNY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[isobel16]

Hi there,

Thank you for that, that is helpful. I am attempting to retrain the HMM using signal files from my population, however I am struggling given that there are 6 values for LRR reflecting 6 states and only 5 for BAF. Could you advise on this please?

[isobel16]

Hi there,

Apologies, just chasing this up as struggling to make any progress, and I see there is a warning on the --train option in PennCNV to train the hmm.