Repository navigation
Error during detector training #73
Description
hello
i faced a problem during detector o my sample
here is the error message, it says i didn't use a negative dataset but actually i used one that is called GeneSwap_Negatives.pfam.tsv
i think deepbgc can't see the negative dataset because of a error o my command. --help didn't tell where or how to put it
'optimizer': 'adam',
'shuffle': True,
'timesteps': 256,
'validation_size': 0,
'verbose': 1,
'weighted': True},
'input_params': { 'features': [ {'type': 'ProteinBorderTransformer'},
{ 'type': 'Pfam2VecTransformer',
'vector_path': 'pfam2vec.csv'}]},
'type': 'KerasRNN'}
INFO 15/05 09:28:22 Loaded 41102 samples and 80777 domains from sample1_deepbgc_prepare_result.tsv
INFO 15/05 09:28:28 Loaded 10128 samples and 706950 domains from GeneSwap_Negatives.pfam.tsv
ERROR 15/05 09:28:33 Got target variable with only one value {0} in: ['sample1_deepbgc_prepare_result.tsv', 'GeneSwap_Negatives.pfam.tsv']
Traceback (most recent call last):
File "/root/esraa/miniconda3/envs/deepbgcv0.1.29/lib/python3.7/site-packages/deepbgc/main.py", line 113, in main
run(argv)
File "/root/esraa/miniconda3/envs/deepbgcv0.1.29/lib/python3.7/site-packages/deepbgc/main.py", line 102, in run
args.func.run(**args_dict)
File "/root/esraa/miniconda3/envs/deepbgcv0.1.29/lib/python3.7/site-packages/deepbgc/command/train.py", line 60, in run
train_samples, train_y = util.read_samples(inputs, target)
File "/root/esraa/miniconda3/envs/deepbgcv0.1.29/lib/python3.7/site-packages/deepbgc/util.py", line 574, in read_samples
'Did you provide positive and negative samples?')
ValueError: ("Got target variable with only one value {0} in: ['sample1_deepbgc_prepare_result.tsv', 'GeneSwap_Negatives.pfam.tsv']", 'At least two values are required to train a model. ', 'Did you provide positive and negative samples?')
ERROR 15/05 09:28:33 ================================================================================
ERROR 15/05 09:28:33 DeepBGC failed with ValueError: Got target variable with only one value {0} in: ['sample1_deepbgc_prepare_result.tsv', 'GeneSwap_Negatives.pfam.tsv']
ERROR 15/05 09:28:33 ================================================================================
ERROR 15/05 09:28:33 At least two values are required to train a model.
ERROR 15/05 09:28:33 Did you provide positive and negative samples?
ERROR 15/05 09:28:33 ================================================================================
my cmmand:
deepbgc train --model templates/deepbgc.json --output MyDeepBGCDetector.pkl sample1_deepbgc_prepare_result.tsv GeneSwap_Negatives.pfam.tsv --config PFAM2VEC pfam2vec
.csv -v ClusterFinder_Annotated_Contigs.full.gbk
Activity
prihoda commentedon May 16, 2022
Hi @esraagithub, your file
sample1_deepbgc_prepare_result.tsvcontains the BGC samples, is that correct? This file will need to contain anin_clustercolumn, which will have a value of1in all rows (in case the file only contains "positive" BGC samples). Your file should also contain asequence_idcolumn which should contain an identifier of each BGC.esraagithub commentedon May 16, 2022
@prihoda
Thank you for your response
Yes this file sample1_deepbgc_prepare_result.tsv resulted from deepbgc prepare command. It actually contain sequence id column and in_cluster column but in_cluster column has 0 in all raws not 1
I don't know why it has only zero
prihoda commentedon May 21, 2022
Hi @esraagithub if that file contains just BGC samples, you can manually change the value to 1 in all rows
esraagithub commentedon May 21, 2022
esraagithub commentedon May 23, 2022
thank you, i tried it and it worked well but i get another error in the next step "training the classifier"
raise ValueError('No overlap found between classes and samples. Classes should be indexed by sequence_id.')
ValueError: No overlap found between classes and samples. Classes should be indexed by sequence_id.
ERROR 23/05 23:53:18 ================================================================================
ERROR 23/05 23:53:18 DeepBGC failed with ValueError: No overlap found between classes and samples. Classes should be indexed by sequence_id.
ERROR 23/05 23:53:18 ================================================================================
I have a "sequence id" column in sample file (which i got from deepbgc prepare ) but no overlaps between it and classes file. so what should i do in this case? is this means i can't proceed with training?
@prihoda