igor vs sonnia + nucleotidic junction

[Eglantine-Hr]

Dear @giulioisac @zacmon ,

thank you for the nice tool.

I am used to Igor so it is already incorparated in a pipeline.

I wanted to have access to Ppost as well, but I have very different values for Pgens computed with Sonnia and Igor.
Is there a major update between both tools?
I wondered if I could use nucleotidic sequences to compute Pgen with sonnia as with Igor.

Here are the versions and command lines

Igor (v1.4.0)

echo "$current_seq"
  if PGEN=$(timeout 600s igor-compute_pgen ${ORGANISM} beta ${current_seq:1:-1}); then
    PGEN=${PGEN#*$'\n'} 
    echo "${current_seq},  ${PGEN}" >> $OUTPUT

where current seq corresponds to AIRR "junction" nucleotidic sequence

Sonnia (latest)
sonnia evaluate --model {params.specie}_T_beta -i {input.sonnia_input} -o {output} --no-header --delimiter " "
where sonnia input is a txt file with junction_aa v_call and j_call from AIRR tsv

A small example is attached (inputs seqeuences are drawn the same airr file)
Sorry if I missed something in the documentation.

Kind regards,

smallexample.txt

[DanielWongPGH]

a quick thought: sonnia is outputting the pgen of the corresponding amino acid sequence, which is the sum of the probabilities of (in principle thousands of) degenerate nucleotide sequences. whereas igor is outputting the pgen of only one such nucleotide sequence. maybe this explains the discrepancy?

[giulioisac]

Yes exactly! For the moment, SoNNia works only in amino acid space, so you will get pgen and ppost of amino acid sequences. If you are really interested in the nucleotide, you can also look at OLGA