LncADeep

An ab initio lncRNA identification and functional annotation tool based on deep learning

Long noncoding RNAs (lncRNAs) play important biological roles and have been implicated in human diseases. To characterize lncRNAs, identifying and annotating lncRNAs is necessary. Here, we propose a novel lncRNA identification and functional annotation tool named LncADeep. First, LncADeep identifies lncRNAs by integrating sequence intrinsic and homology features based on deep belief networks. Second, LncADeep predicts lncRNA-protein interactions using sequence and structure features based on deep neural networks. Third, since accurate lncRNA-protein interactions can help to infer the functions of lncRNAs, LncADeep conducts KEGG and Reactome pathway enrichment analysis and functional module detection with the predicted interacting proteins of lncRNAs. Case studies show that LncADeep's annotations for lncRNAs comply with their known functions. As a tool for lncRNA identification and functional annotation based on deep learning, LncADeep has outperformed state-of-the-art tools on predicting lncRNAs and lncRNA-protein interactions, and can automatically provide informative functional annotations for lncRNAs.

Version

• LncADeep 1.0 (Tested on Linux_64, including CentOS 6.5 and Ubuntu 16.04)

Download

• Release version: LncADeep_v1.0

• Human test dataset for lncRNA identification.

• Mouse test dataset for lncRNA identification.

• Functional annotations for lncRNAs in case studies.

• Functional annotations for the lncRNAs collected in GENCODE v24.

Citation

Cheng Yang, Longshu Yang, Man Zhou, Haoling Xie, Chengjiu Zhang, May D Wang, Huaiqiu Zhu; LncADeep: An ab initio lncRNA identification and functional annotation tool based on deep learning, Bioinformatics, 2018, bty428

Contact

Please direct your questions to: Dr. Huaiqiu Zhu, hqzhu@pku.edu.cn

LncADeep.py

For lncRNA identification.

• Python version >= 2.7.6

• numpy version >= 1.8.2

• HMMER package 3.1b2

• Pfam database 29.0

For predicting lncRNA-protein interactions and annotating lncRNA functions.

• Python, version >= 2.7.6

• numpy, version >= 1.8.2

• pandas, version >= 0.18.0

• theano, version >= 0.8.2

• Keras, version 1.2.2

• h5py, version >= 2.5.0

• MCL package

• R, version >= 3.3.2

• iGraph R package

Prerequisites

Please install numpy, theano, pandas, Keras, h5py, and iGraph according to their manuals. The following are examples for installing these prerequisites.

numpy, theano, pandas, and h5py are python packages, which can be installed with pip, for example:

# we use python v2.7.13
# our machine is implemented with the following versions

pip install numpy  # numpy v1.13.1
pip install Theano  # Theano v0.9.0
pip install pandas  # pandas v0.20.3
pip install h5py   # h5py v2.7.0

iGraph is an R package, which can be installed with:

# we use R v3.3.2
# Download and install the package

install.packages("igraph")

Keras is a Python Deep Learning library. For Keras, please be noted that we use Keras v1.2.2, and we use theano as its backend. Please edit the file ~/.keras/keras.json and change the backend. For example,

# Download Keras v1.2.2
wget https://github.com/fchollet/keras/archive/1.2.2.tar.gz

# unpack the zipped file
tar xzvf 1.2.2.tar.gz

# install Keras v1.2.2
cd keras-1.2.2
python setup.py install

# edit ~/.keras/keras.json and change the backend
{    
    "image_dim_ordering": "th",
    "epsilon": 1e-07,
    "floatx": "float32",
    "backend": "theano"
}

HMMER and MCL package have been included in LncADeep package. You don't need to install them yourself. After installing the above prerequisites, you can now install LncADeep.

Install LncADeep from zipped file

download the zipped file

wget http://cqb.pku.edu.cn/ZhuLab/LncADeep/LncADeep_v1.0.tgz

unpack the zipped file

tar xzvf LncADeep_v1.0.tgz

change directory to LncADeep

cd LncADeep_v1.0

configure and add directory to the PATH, and you are done!

chmod +x configure
./configure

source $HOME/.bash_profile

Install LncADeep using git

clone LncADeep package

git clone https://github.com/cyang235/LncADeep.git

change directory to LncADeep

cd LncADeep

download Pfam 29.0 database

wget ftp://ftp.ebi.ac.uk/pub/databases/Pfam/releases/Pfam29.0/Pfam-A.hmm.gz 
gzip -d Pfam-A.hmm.gz
mv Pfam-A.hmm ./LncADeep_lncRNA/src/

# the Pfam-A.hmm need be put in directory /path to LncADeep/LncADeep_lncRNA/src/

configure and add directory to the PATH, and you are done!

chmod +x configure
./configure

source $HOME/.bash_profile

LncADeep.py

An ab initio lncRNA identification and functional annotation tool based on deep learning

usage: 

An ab initio lncRNA identification and functional annotation tool based on deep learning

optional arguments:
  -h, --help       show this help message and exit
  -v, --version     show program's version number and exit   
  -MODE {lncRNA,anno},--MODE {lncRNA,anno}
              (Required) The mode used for lncRNA identification or
              functional annotation. If  is chosen, LncADeep
              will identify lncRNAs. If  is chosen, LncADeep
              will predict lncRNA-protein interactions and annotate
              lncRNA functions. Default is 
  -o OUT_PREFIX,--out OUT_PREFIX
              (Required) The output prefix of results
  -f FASTA_FILE,--fasta FASTA_FILE
              (Required for lncRNA identification) Sequence file in
              FASTA format to be predicted
  -m {full,partial}, --model {full,partial} 
              (Optional for lncRNA identification) The model used 
              for lncRNA identification, default is 
  -s {human,mouse},--species {human,mouse} 
              (Optional for lncRNA identification) The species used 
              for lncRNA identification, default is 
  -th THREAD,--thread THREAD 
              (Optional for lncRNA identification) Use multi-thread 
              for predicting, default is 1
  -HMM HMMTHREAD,--HMMthread HMMTHREAD 
              (Optional for lncRNA identification) The thread number
              of using HMMER, default is 8
  -l RNA_FILE, --lncRNA RNA_FILE
              (Required for functional annotation) lncRNA sequence 
              file in FASTA format
  -p PROTEIN_FILE,--protein PROTEIN_FILE 
              (Optional for functional annotation) protein sequence 
              file in FASTA format
  -a {1,0},--annotation {1,0} 
              (Optional for functional annotation) To annotate
              lncRNA functions. If  is selected, LncADeep will
              annotate the functions for lncRNAs, otherwise LncADeep
              will only give the interacting proteins for lncRNAs.
              The default is .
  -r PAIR_FILE,--pair PAIR_FILE 
              (Optional for functional annotation) The lncRNA-
              protein pairs to be predicted. If this option is
              selected, LncADeep will only output interacting
              proteins.

---

Built with MkDocs using a theme provided by Read the Docs.