Log-transform expr? will it be log 10 or log2

50 views
Skip to first unread message

Karl Clauser

unread,
Jan 25, 2025, 11:19:34 AM1/25/25
to HLAthena
Hi Sisi,

When selecting the option Log-transform expr? 
http://hlathena.tools/         predict-> How to
suggests that HLAthena will perform log10 (TPM+1)

I often have RNAseq expression in the form of log2(TPM) with 0 set to NA.
If I input this to HLAthena and set logtransform_expr to false, what would you expect to happen internally with HLAthena?
Would this work sensibly for HLAthena MSiCE as it provides relative expression?

What would you advise that I do?

Thanks,

--Karl

Sisi Sarkizova

unread,
Jan 25, 2025, 8:17:54 PM1/25/25
to Karl Clauser, HLAthena
Hi Karl, thanks for the question!

For accurate prediction with MSiCE, I would suggest to undo the log2 TPM transform (with exponentiation) and then pass the original TPM values to HLAthena and specify log-tranform = True. 

If the log2 values are passed in with log-transform=False, the distribution of TPM values won't correspond to the distribution of the learned data, i.e. the log2 values would be an 'overestimate' as compared to log10 and thus the MSiCE predictions would be higher than expected. The cases where the TPM=0 are handled via the +1 in the built-in log-transform option.

I hope this helps!
Sisi


--
You received this message because you are subscribed to the Google Groups "HLAthena" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hlathena+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/hlathena/eb7e2863-0912-458b-a6e3-292253110cd8n%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages