Welcome to Scikit-ribo’s documentation!

_images/logo.png

Contents:

Getting Started

This document will show you how to install and run Scikit-ribo.

What is Scikit-ribo

Scikit-ribo is an open-source software for accurate genome-wide A-site prediction and translation efficiency inference from Riboseq and RNAseq data.

Source Code: https://github.com/hanfang/scikit-ribo

Introduction

Scikit-ribo has two major modules:

  • Ribosome A-site location prediction using random forest with recursive feature selection
  • Translation efficiency inference using a codon-lvel generalized linear model with ridge penalty

A complete analysis with scikit-ribo has two major procedures:

  • The data pre-processing step to prepare the ORFs, codons for a genome: scikit-ribo-build.py
  • The actual model training and fitting: scikit-ribo-run.py

Detailed workflow

_images/methods.png

Inputs

  • The alignment of Riboseq reads (bam)
  • Gene-level quantification of RNA-seq reads (from either Salmon or Kallisto)
  • A gene annotation file (gtf)
  • A reference genome for the model organism of interest (fasta)

Output

  • Translation efficiency estimates for the genes
  • Translation elongation rate for 61 sense codons
  • Ribosome profile plots for each gene
  • Diagnostic plots of the models

Cite

Fang et al, “Scikit-ribo: Accurate inference and robust modelling of translation dynamics at codon resolution” (Preprint coming up)

Contact

Han Fang

Stony Brook University & Cold Spring Harbor Laboratory

Email: hanfang.cshl@gmail.com

Requirement

Environment

  • Python3
  • Linux
  • Recommend setting up your environment with Conda

Dependencies

  • Command-line pacakges:
Python package Version >=
bedtools 2.26.0
  • Python package:
Python package Version >=
colorama 0.3.7
glmnet_py 0.1.0b
gffutils 0.8.7.1
matplotlib 1.5.1
numpy 1.11.2
pandas 0.19.2
pybedtools 0.7.8
pyfiglet 0.7.5
pysam 0.9.1.4
scikit_learn 0.18
scipy 0.18.1
seaborn 0.7.0
termcolor 1.1.0

Note: When using pip install scikit-ribo, all the following dependencies will be pulled and installed automatically.

Installation

Options

There are three options to install Scikit-ribo.

  1. Install Scikit-ribo with pip:

    pip install scikit-ribo

  2. Install Scikit-ribo with conda/biocodon:

    Coming up

  3. Compile from source:

    git clone https://github.com/hanfang/scikit-ribo.git
cd scikit-ribo
python setup.py install

Test whether the installation is successful

Once the installation is successful, you should expect the below if you type:

scikit-ribo-run.py
_images/successful_installation.png

Usage

How-to-use

  • Pre-process: scikit-ribo-build.py
  • Fit model: scikit-ribo-run.py

Detailed usage

  • Pre-process: scikit-ribo-build.py:

    scikit-ribo-build.py -g gtf-file -f fasta-file -p prefix -r rna-fold-file -t TPM-file -o index-path

required arguments:

-g G        Gtf file, required
-f F        Fasta file, required
-p P        Prefix to use, required
-r R        Rnafold file, required
-t T        TPM of RNAseq sample, required
-o O        Output path of the built indexes, required

  • Fit model: scikit-ribo-run.py:

    scikit-ribo-run.py -i bam-file -f index-path -p prefix -o output-path

required arguments:

-i I        Input bam file, required
-f F        path to the Folder of BED/index files generated by the pre-processing module, required
-p P        Prefix for BED/index files, required
-o O        Output path, recommend using the sample id, required

optional arguments:

-h, --help  show this help message and exit
-q Q        minimum mapQ allowed, Default: 20
-s S        Shortest read length allowed, Default: 10
-l L        Longest read length allowed, Default: 35
-c          enable cross validation for glmnet
-r          setting this flag will enable the RelE mode
-u U        Un-mappable regions

For more information, please refer to the template shell script about details of executing the two modules.

Preparing Input Data

  • RNAfold: call_rnafold.py:

    python call_rnafold.py  -f <prefix>.expandCDS.fasta -r rnafold-binary -p prefix -n num-processes -o output-folder

Resulting file: <output-folder>/<prefix>.rnafold_lbox.txt Pre-built files can be also downloaded from here: https://github.com/hanfang/scikit-ribo/tree/master/data/prebuilt_rnafold

  • TPM: Gene-level quatification from Salmon or Kallisto

Output format

  • Translation efficiency estimates for the genes: genesTE.csv
gene log2_TE
YAL001C -0.7444
YAL002W -0.9811
YAL003W 2.0833
... ...
  • Translation elongation rate for 61 sense codons: codons.csv
codon codon_dwell_time
AAA 0.9795
AAC -0.9811
... ...
TTT 2.0833
  • Diagnostic plots of the models
  1. asite_feature_importances.pdf: Feature importance plot for the random forest classifier.
  2. asite_roc.pdf: ROC curve for the random forest classifier.
  3. asite_3offset.pdf: Distribution of A-site by read length and 3’ phase
  4. asite_5offset.pdf: Distribution of A-site by read length and 5’ pahse
  5. (optional) riboseq.lambda_cv.pdf: If cross-validation is enabled, the cross-validation curve is plotted.

Contact

Support

If you are having issues, please let us know: hanfang.cshl@gmail.com

License

The project is licensed under the GPLv2 license.

Developer

Han Fang

Stony Brook University & Cold Spring Harbor Laboratory

Email: hanfang.cshl@gmail.com

Cite

Fang et al, “Scikit-ribo: Accurate inference and robust modelling of translation dynamics at codon resolution” (Preprint coming up)

Indices and tables