Sequences Invalid

[ciancone94]

Hello,

I was trying to batch some sequences, some with multiple mods, and I got an error in the command line text output that said:

collected 4 valid samples, discard 2 invalid samples

with no additional information about why the sequences were rejected.

These same sequences were accepted without any Mox/Ccm mods - attached is the input .txt file I used. 'synthetic1' and 'synthetic2' were the ones rejected and I just copied them and re-tried without mods and they ran just fine. I'd prefer to be able to run them with Ccm always and Mox if possible (although this one is less important).

Side question - is the way I'm declaring multiple mods correct? I tried following the syntax on the main page, but I didn't see an example for multiple mods (I may have missed this, apologies if so).

Any help is appreciated, thanks!

-Anthony
pDeepXL_sequences.txt

[NorthblueM]

Hi,

Due to the limited number of samples with a charge state of 6 or higher, pDeepXL does not support these charge states during training. Internally, the code restricts allowed charge states to the range [2, 5], as specified in ./configs/config.ini.

After reviewing the file you provided, pDeepXL_sequences.txt, this limitation could explain why you are encountering invalid samples.

-Pengzhi

[ciancone94]

Hello,

Thank you very much, this does appear to solve one issue; however, whenever I try adding multiple mods to a peptide sequence, I get the following traceback error:

loading predicting data...
0%| | 0/6 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/home/cianconeam/pDeepXL_batch_test1_250113.py", line 9, in
predictions=pDeepXL.predict.predict_batch(DATA_FILE, False)
File "/home/cianconeam/anaconda3/lib/python3.10/site-packages/pDeepXL/predict.py", line 206, in predict_batch
predict_data_loader,mpTitleLines,header=LoadPredictDataSet(BATCH_SIZE, is_non_cleavable, path_data_file)
File "/home/cianconeam/anaconda3/lib/python3.10/site-packages/pDeepXL/predict.py", line 100, in LoadPredictDataSet
T,PT,L,L1,L2,S1,S2,X=GetPeptideFeatures(CSMs, consider_xlink_ion)
File "/home/cianconeam/anaconda3/lib/python3.10/site-packages/pDeepXL/predict.py", line 74, in GetPeptideFeatures
seq1_vec,seq1_aa_mod_sum = pf.Sequence2Vec(seq1,mods1,prec_charge,instrument,NCE_low, NCE_medium, NCE_high,LinkerName,linksite1,False)
File "/home/cianconeam/anaconda3/lib/python3.10/site-packages/pDeepXL/featurer.py", line 98, in Sequence2Vec
if i in mod:
TypeError: 'in ' requires string as left operand, not int

The problematic input looks like this: {1: 'Oxidation[M]', 6: 'Oxidation[M]', 7: 'Carbamidomethyl[C]'}

I was getting this error with the initial input, so I tried changing how the mods looked according to the example in featurer.py:

seq_vec, left_sum, right_sum = pf.Sequence2Vec('MIAKSEQEIGK',{0:'Oxidation[M]',5:'Carbamidomethyl[C]',7:'Oxidation[M]'},3,'Lumos',0,30,0,'DSS',3,False)

This still resulted in the same TypeError. I tried reformatting the input as {1,7:'Oxidation[M]','Carbamidomethyl[C]'} but this also results in a type error. Any way to add multiple mods to the peptide?

Thanks,

Anthony

[NorthblueM]

Before line 98 if i in mod, in Sequence2Vec within featurer.py, You can try to add the following debug statements:

print(type(mod))
print(mod)

Here, mod needs to be dictionary type. You can verify if mod is a dictionary and matches the modifications you enterend, such as {1: 'Oxidation[M]', 6: 'Oxidation[M]', 7: 'Carbamidomethyl[C]'}

The pDeepXL.predict.predict_batch function reads file through the following chain of methods predict.LoadPredictDataSet -> utils.ReadpLink2PredictFile -> utils.ReadpLink2MatchInfoFile. In utils.py line 608, you can observe that the modification string "{1: 'Oxidation[M]', 6: 'Oxidation[M]', 7: 'Carbamidomethyl[C]'}" in the text will convert into dictionary type. The keys are of type int and the values are of type string.

Feel free to continue if you have any questions.
-Pengzhi