Running Example, pandas/pickle issues

[CFGrote]

Hi there,
first. thanks for making your code accessible, this is very useful.

I run into the problem detailed below when running the example from the README.

OS / python env setup:

• OS: ubuntu 18.04
• env: conda with cudatoolkit=9.0, cudnn, python=3.6 and all dependencies according to requirements.txt

deepgoplus installation:

• clone deepgoplus repo, on commit c309ade
• installed deepgoplus with

python setup.py install

Download data

$> wget https://deepgo.cbrc.kaust.edu.sa/data/data.tar.gz 

and unpack into data/data directory.

Get a test sequence from training data

$> head -2 data/data/test_data.fa > test_single.fa                                           
$> cat test_single.fa                                                                          
>KR103_HUMAN
MATSTMSVCSSAYSDSWQVDACPESCCEPPCCATSCCAPAPCLTLVCTPVSCVSSPCCQAACEPSPCQSGCTSSCTPSCCQQSSCQPACCTSSPCQQACCVPVCCKPVCCVPVCCKPVCCKPICCVPVCSGASSSCCQQSSRQPACCTTSCCRPSSSVSLLCRPVCRSTCCVPIPSCCAPASTCQPSCCRPASCVSLLCRPTCSRLSSACCGLSSGQKSSC

Run deepgoplus (here's where the problems start)

• run deepgoplus:

$> deepgoplus --data-root data/data --in-file test_single.fa                                  [master] 
Traceback (most recent call last):
  File "XXX/lib/python3.6/site-packages/pandas/io/pickle.py", line 166, in try_read
    return read_wrapper(lambda f: pkl.load(f))
  File "XXX/lib/python3.6/site-packages/pandas/io/pickle.py", line 149, in read_wrapper
    return func(f)
  File "XXX/lib/python3.6/site-packages/pandas/io/pickle.py", line 166, in <lambda>
    return read_wrapper(lambda f: pkl.load(f))
ValueError: unsupported pickle protocol: 5

• I assume the pickle files were written with pickle protocol 5, convert to protocol 4:
  In a different python environment (with pandas version 1.0.5)

import pandas, glob
for pkl in glob.glob('data/data/*.pkl'):
    pandas.read_pickle(pkl).to_pickle(pkl, protocol=4)

• This gets rid of the error but another one now comes up (switch back to py3.6/pandas0.23 env)

$> deepgoplus --data-root data/data --in-file test_single.fa                                  [master] 
Traceback (most recent call last):
  File "XXX/bin/deepgoplus", line 42, in <module>
    sys.exit(load_entry_point('deepgoplus==1.0.0', 'console_scripts', 'deepgoplus')())
  File "XXX/lib/python3.6/site-packages/click/core.py", line 764, in __call__
    return self.main(*args, **kwargs)
  File "XXX/lib/python3.6/site-packages/click/core.py", line 717, in main
    rv = self.invoke(ctx)
  File "XXX/lib/python3.6/site-packages/click/core.py", line 956, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "XXX/lib/python3.6/site-packages/click/core.py", line 555, in invoke
    return callback(*args, **kwargs)
  File "XXX/lib/python3.6/site-packages/deepgoplus-1.0.0-py3.6.egg/deepgoplus/main.py", line 63, in main
  File "XXX/lib/python3.6/site-packages/pandas/core/frame.py", line 2668, in __getitem__
    is_mi_columns = isinstance(self.columns, MultiIndex)
  File "XXX/lib/python3.6/site-packages/pandas/core/generic.py", line 4372, in __getattr__
    return object.__getattribute__(self, name)
  File "pandas/_libs/properties.pyx", line 65, in pandas._libs.properties.AxisProperty.__get__
  File "XXX/lib/python3.6/site-packages/pandas/core/generic.py", line 4372, in __getattr__
    return object.__getattribute__(self, name)
AttributeError: 'DataFrame' object has no attribute '_data'

• At this point I gave up. Have the requirements changed but documentation is not updated? Any help would be greatly appreciated, thanks in advance!

[CFGrote]

I found the following fix:

• checkout PR Bump tensorflow-gpu from 1.12.0 to 2.3.1 #28
• remove all version pinning in requirements.txt except for "tensorflow-gpu==2.3.1"
• upgrade python to version 3.7
• not install deepgoplus through pip or setup.py but use

python deepgoplus/main.py 

instead.

• upgrade diamond db following the solution given in diamond version Error #32

[coolmaksat]

Hi,
Thank you for raising this issue and PR.
We will update the master branch soon.