bhi-kimlab/deepfam

Deep learning based alignment-free method for protein family modeling and prediction

About DeepFam

DeepFam is a deep learning based alignment-free protein function prediction method. DeepFam first extracts features of conserved regions from a raw sequence by convolution layer and makes a prediction based on the features.

Features

• Alignment-free: Do not need multiple or pairwise sequence alignment to train family model.
• Instead, locally conserved regions within a family are trained by convolution units and 1-max pooling. Convolution unit works similar as PSSM.
• Utilizing variable-size convolution unit (multiscale convolution unit) to train family specific conserved regions whose lengths are usually various.

Installation

DeepFam is implemented in with Tensorflow library. Both CPU and GPU machines are supported. For detail instruction of installing Tensorflow, see the guide on official website.

Requirements

• Python: 2.7
• Tensorflow: over 1.0

Usage

First, clone the repository or download compressed source code files.

$ git clone https://github.com/bhi-kimlab/DeepFam.git
$ cd DeepFam

You can see the valid paramenters for DeepFam by help option:

$ python src/DeepFam/run.py --help

One example of parameter setting is like:

$ python src/DeepFam/run.py \
  --num_windows [256, 256, 256, 256, 256, 256, 256, 256] \ 
  --window_lengths [8, 12, 16, 20, 24, 28, 32, 36] \
  --num_hidden 2000 \
  --batch_size 100 \
  --keep_prob 0.7 \ 
  --learning_rate 0.001 \
  --regularizer 0.001 \ 
  --max_epoch 25 \
  --seq_len 1000 \ 
  --num_classes 1074 \ 
  --log_interval 100 \ 
  --save_interval 100 \ 
  --log_dir '/tmp/logs' \  
  --test_file '/data/test.txt' \ 
  --train_file '/data/train.txt' 

Data

All data used by experiments described in manuscript is available at here.

Contact

If you have any question or problem, please send a email to [email protected]

deepfam's Issues

How do I use this tool?

Hi there,

probably a stupid question, but... how do I use this tool?
I see that I can train a model and test it on some other data.
But is there any option to save the model and get random data predicted?
We have a big database and would need to build 400+ models (if we'd be using it), and would like to annotate then new proteins iteratively. But I don't see how this would work here.

Any hints please :)?
Thanks

data?

Hello,

It seems I can't access the page for the data. Could you please update the link?

Thank you!