search

Arcadia-Science / peptigate

Peptigate ("peptide" + "investigate") predicts bioactive peptides from transcriptome assemblies or sets of proteins.
MIT License
0 stars 0 forks source link

report plmutils and deeppeptide scores in the predictions/peptide_predictions.tsv output file #48

Open taylorreiter opened 3 months ago

taylorreiter commented 3 months ago

Description of feature

The predictions/peptide_predictions.tsv file looks like this right now:

peptide_id      start   end     peptide_type    peptide_class   prediction_tool nlpprecursor_class_score        nlpprecursor_cleavage_score     protein_sequence        nucleotide_sequence
Transcript_100036.p2_start55_end96      55      96      cleavage        LASSO_PEPTIDE   nlpprecursor    0.5960474014282227      -19422.001953125        RSVAEGTTLTPWKERKKAAAIVFASKRFPHLSAHSFLLPPP       CGTT>
Transcript_1000448.p1_start91_end131    91      131     cleavage        CLASS_I_LANTIPEPTIDE    nlpprecursor    0.9570119380950928      -19379.048828125        SPPIDGCEGYVRSNETCLTLNCKSRGKTCCNGPCGDPFCV    >
Transcript_100028.p1_start101_end112    101     112     cleavage        Propeptide      deeppeptide     NA      NA      TAGLHVSAIAPV    ACGGCCGGCCTTCACGTCAGCGCCATTGCACCTGTT`

The only score we report is for nlpprecursor. Both DeepPeptide and plm-utils also have scores, so we could add a column reporting these as well

taylorreiter commented 3 months ago

The plmutils part is completed by #52. DeepPeptide will take a little more engineering because those scores are in a json and not currently extracted I don't think.