pybedtools error

[BastiKarg]

executing the testcases results in the following error: Traceback (most recent call last): File "train.py", line 301, in main() File "train.py", line 197, in main utils.load_chip_singleTask(input_dirs, tf) File "/home/hpc/t1172/di36dih/bin/FactorNet-master/utils.py", line 462, in load_chip_singleTask blacklist = make_blacklist() File "/home/hpc/t1172/di36dih/bin/FactorNet-master/utils.py", line 373, in make_blacklist blacklist = blacklist.cat(blacklist2) File "/home/hpc/t1172/di36dih/anaconda2/envs/env1/lib/python2.7/site-packages/pybedtools/bedtool.py", line 840, in decorated result = method(self, *args, **kwargs) File "/home/hpc/t1172/di36dih/anaconda2/envs/env1/lib/python2.7/site-packages/pybedtools/bedtool.py", line 3064, in cat for f in self: File "pybedtools/cbedtools.pyx", line 798, in pybedtools.cbedtools.IntervalIterator.next File "pybedtools/cbedtools.pyx", line 698, in pybedtools.cbedtools.create_interval_from_list pybedtools.cbedtools.MalformedBedLineError: Unable to detect format from ['chr17_ctg5_hap1', '-1', '1255794']

Even after updating pybedtools or changing of the genome the error still occurs.

[BastiKarg]

Update: I managed to work around this error by removing all non regular chrs from the blacklist.bed.gz.

Further trying to execute the example train.py led me to other errors in the model building step though:

33885000 training samples 27862 validation samples Building model Traceback (most recent call last): File "train.py", line 301, in main() File "train.py", line 258, in main model = utils.make_model(num_tfs, num_bigwigs, num_motifs, num_recurrent, num_dense, dropout_rate) File "/naslx/projects/t1172/di36dih/FactorNet-master/utils.py", line 572, in make_model MaxPooling1D(stride=13, pool_size=13), # strides -> stride File "/home/hpc/t1172/di36dih/anaconda2/envs/env1/lib/python2.7/site-packages/keras/layers/pooling.py", line 75, in init border_mode, **kwargs) File "/home/hpc/t1172/di36dih/anaconda2/envs/env1/lib/python2.7/site-packages/keras/layers/pooling.py", line 16, in init super(_Pooling1D, self).init(**kwargs) File "/home/hpc/t1172/di36dih/anaconda2/envs/env1/lib/python2.7/site-packages/keras/engine/topology.py", line 307, in init assert kwarg in allowed_kwargs, 'Keyword argument not understood: ' + kwarg AssertionError: Keyword argument not understood: pool_size

These are my current versions:

Keras 1.1.1 Keras-Applications 1.0.7 Keras-Preprocessing 1.0.9 tensorflow 1.5.1 tensorflow-estimator 1.14.0 tensorflow-tensorboard 1.5.1 torch 1.1.0

What are the recommended versions of keras, tensorflow, etc.? Currently I am using keras=1.1.1 (as I greped the version out of https://github.com/uci-cbcl/FactorNet/blob/bef6f6b38e81d362162a106dc8a726ecae910138/models/E2F1/onePeak_Unique35_DGF/model.json) but could not figure out which tensorflow version suits accordingly https://pypi.org/project/tensorflow/#history

[jhawe]

Not sure this is an issue with FactorNet, I feel like it is just really picky when it comes to the versions of the underlying software. I managed to create a conda environment with specific library versions which seems to do the trick for me (initially having similar errors as @TripelB , also filtered non standard chromosomes from the blacklist files).

Below is the conda env I used to get FactorNet to run:

name: factornet

channels:
  - conda-forge
  - bioconda
  - defaults
  - pytorch
  - anaconda

dependencies:
  - python =2.7.15
  - mkl
  - mkl-service
  - theano =0.9.0
  - pyfasta
  - pybigwig =0.2.8
  - pybedtools
  - parmap
  - keras =1.2.2
  - cuda80
  - cudnn
  - numpy =1.10.4
  - scipy =0.17.0
  - matplotlib

Maybe this helps someone else, too.

EDIT: I updated the yaml to be more specific for pyBigWig (which failed on another system where a different version was obtained via conda) and added matplotlib in order to be able to run the predict.py