Paper abstract / Intro:

Accurate annotation of protein functions is important for a profound understanding of molecular biology. A large number of proteins remain uncharacterized because of the sparsity of available supporting information. For a large set of uncharacterized proteins, the only type of information available is their amino acid sequence. In this paper, we propose DeepSeq -- a deep learning architecture -- that utilizes only the protein sequence information to predict its associated functions. The prediction process does not require handcrafted features; rather, the architecture automatically extracts representations from the input sequence data. Results of our experiments with DeepSeq indicate significant improvements in terms of prediction accuracy when compared with other sequence-based methods. Our deep learning model achieves an overall validation accuracy of 86.72%, with an F1 score of 71.13%. Moreover, using the automatically learned features and without any changes to DeepSeq, we successfully solved a different problem i.e. protein function localization, with no human intervention. Finally, we discuss how this same architecture can be used to solve even more complicated problems such as prediction of 2D and 3D structure as well as protein-protein interactions.

Authors:

• Nauman ([email protected], [email protected], recluze.wordpress.com) -- Queries about ML should go here.
• Hafeez ur Rehman ([email protected]) -- Queries about Bioinformatics should go here.

Preprint of related publication available here: http://www.biorxiv.org/content/early/2017/07/25/168120

Import points:

- Requires python2.7
- See requirements.txt for exact version of libraries used. Keras v1.2.1 gives errors so use keras 1.2.0
- I've used theano backend for keras. If you use tensorflow, I think you will have issues.
- It's suggested that you use `virtualenv` to create a new environment and then install required packages.
    ```
    pip install virtualenv
    virtualenv bi
    cd bi
    . bin/activate
    git cone <git_repo_url>
    pip install -r <git_repo_name>src/requirements.txt
    ```
- Set keras/theano to use the GPU. (Only do this on the GPU machine.) Put the following in `~/.theanorc`
    ```
    [global]
    device = gpu
    floatX = float32
    ```
- Set keras to use theano. In `~/.keras/keras.json`, put the following:
    ```
    {
        "image_dim_ordering": "th",
        "epsilon": 1e-07,
        "floatx": "float32",
        "backend": "theano"
    }
    ```

Execution

The source is executed in several steps.

1. First, data needs to be downloaded to data-scrapes folder.

   • Needs to be in FASTA format along with annotation file in .txt
   • This is already done for human proteins

2. These scrapes need to be converted to a format that we read later.

   • This is done through the python src/scrape2vec.py.
   • Variables to set: scrape_dir, out_file, out_file_fns, out_file_unique_functions
   • Use function_usage_cutoff variable to remove function used fewer times than this number
   • This step has already been done for human proteins downloaded in step 1.

3. Once output files are created from above step, you can run training/validation.

   • This is done through python train.py
   • Some variables need to be set (although current train.py can be executed as is to reproduce our experiments):
     • See top of train.py for parameters of training that you can set
     • target_function can be set to train for a particular function. Set to empty string to train for all functions
     • To quickly check code on slow machines, set restrict_sample_size to, say, 10.
     • results_dir is where results will be stored. These will be -console.txt, -results.txt and -saved-model.h5 prefixed with exp_id i.e. the experiment ID.
     • Bottom of train.py, need to set sequences_file and funtions_file created from step 2 above
     • In utils.py, need to set unique_function_file variable. (Sorry for this clumsiness. I'm too lazy to fix this.)
   • Actual model is defined in get_*_model functions in models.py. This is called from train.py during training.

License

This code is provided under the MIT License.

Copyright 2017 Mohammad Nauman, Hafeez-ur-Rehman

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.