DeepGSR: An optimized deep-learning structure for the recognition of genomic signals and regions

Recognition of different genomic signals and regions (GSRs) in the DNA is helpful in gaining knowledge to understand genome organization and gene regulation as well as gene function. Accurate recognition of GSRs enables better genome and gene annotation. Although many methods have been developed to recognize GSRs, their pure computational identification remains challenging. Moreover, various GSRs usually require a specialized set of features for developing robust recognition models. Recently, deep-learning (DL) methods have been shown to generate more accurate prediction models than the ‘shallow’ methods without the need to develop specialized features for the problems in question. Here, we explore the potential use of DL for the recognition of GSRs.

We developed DeepGSR, an optimized DL architecture for the prediction of different types of GSRs. The performance of the DeepGSR structure is evaluated on the recognition of polyadenylation signals (PAS) and translation initiation sites (TIS) of different organisms: human, mouse, bovine and fruit fly. The results show that DeepGSR outperformed the state-of-the-art methods, reducing the classification error rate of the PAS and TIS prediction in the human genome by up to 29% and 86%, respectively. Moreover, the cross-organisms and genome-wide analyses we performed, confirmed the robustness of DeepGSR and provided new insights into the conservation of examined GSRs across species.

README:

DeepGSR: An optimized deep-learning structure for the recognition of genomic signals and regions. Version 1.1 13/Dec/2017

WHAT IS IT?

-----------

DeepGSR is a deep-learning model that can be used for the recognition of genomic signals and regions with Eukaryotic DNA. It has been applied to polyadenylation signals (PAS) and translation initiation site (TIS).

It uses fasta format DNA Sequences as input. But you can process the data using the provided code.

COMMAND LINE VERSION

--------------------

Here we include the source code of DeepGSR written in Python language and using Keras library with Theano backend.

INSTALLATION

------------

DeepGSR is able to run on any linux platform. To run DeepGSR:

1. Install scikit-learn (http://scikit-learn.org/), keras (https://keras.io/) and cuda for if you want faster processing using GPUs.
2. The data that were used in the paper found in the (Data) folder.
3. There are two types of DeepGSR usage, either for testing using pre-trained models or for training new models; each of these types found in a separate folder.
4. Open a new terminal, then go to the directory that contains the python code. For example: cd Testing/    or   cd Training/DeepGSR-2DCNN
5. Running DeepGSR, command line options: python CNN_Testing.py –h      or        python 2DCNN.py –h

EXAMPLE:

--------

Note: all required data is included in this package

Train DeepGSR on human genome for PAS recognition:

python 2DCNN.py --inputFile ../../Data/Human/PAS_processed/hs_mixAATAAA_polyA.txt --DataName human_AATAAA --FileName human_AATAAA

Train DeepGSR on human genome for TIS recognition:

python 2DCNN.py --inputFile ../../Data/Human/TIS_processed/hs_mixATG_TIS.txt --DataName human_ATG --FileName human_ATG

Test DeepGSR on mouse genome using human trained model for PAS recognition:

python CNN_Testing.py --inputFile ../../Data/Mouse/PAS_processed/mm_mixAATAAA_polyA.txt –inputModel ../human_AATAAA_Model.h5 --DataName mouse_AATAAA --FileName mouse_human_AATAAA

Test DeepGSR on mouse genome using human trained model for TIS recognition:

python CNN_Testing.py --inputFile ../../Data/Mouse/TIS_processed/mm_mixATG_TIS.txt --inputModel ../human_ATG_Model.h5 --DataName mouse_ATG --FileName mouse_human_ATG

CONTACTS

--------

• If you want to report bugs or have general queries email to:
  • manal.kalkatawi@kaust.edu.sa