labranchor

Authors: Paggi J.M., Bejerano

Version: 0.1

License: MIT

Contributed by: Jun Cheng

Cite as: https://doi.org/10.1101/185868

Trained on:

Type: keras

Postprocessing: variant_effects

LaBranchoR (LSTM Branchpoint Retriever) uses a LSTM network built with keras to predict the position of RNA splicing branchpoints relative to a three prime splice site. Precisely evaluating LaBranchoR was challenging due to pervasive noise in the experimental data, but as we show in our paper, we estimate that LaBranchoR correcty predicts a branchpoint for over 90% of 3\'ss. Github link https://github.com/jpaggi/labranchor.

Create a new conda environment with all dependencies installed
kipoi env create labranchor
source activate kipoi-labranchor
Install model dependencies into current environment
kipoi env install labranchor
Test the model
kipoi test labranchor --source=kipoi
Make a prediction
cd ~/.kipoi/models/labranchor
kipoi predict labranchor \
  --dataloader_args='{'fasta_file': 'example_files/hg19.chr22.fa', 'gtf_file': 'example_files/hg19.chr22.gtf', 'length': 70}' \
  -o '/tmp/labranchor.example_pred.tsv'
# check the results
head '/tmp/labranchor.example_pred.tsv'
Get the model
import kipoi
model = kipoi.get_model('labranchor')
Make a prediction for example files
pred = model.pipeline.predict_example()
Use dataloader and model separately
# setup the example dataloader kwargs
dl_kwargs = {'fasta_file': 'example_files/hg19.chr22.fa', 'gtf_file': 'example_files/hg19.chr22.gtf', 'length': 70}
import os; os.chdir(os.path.expanduser('~/.kipoi/models/labranchor'))
# Get the dataloader and instantiate it
dl = model.default_dataloader(**dl_kwargs)
# get a batch iterator
it = dl.batch_iter(batch_size=4)
# predict for a batch
batch = next(it)
model.predict_on_batch(batch['inputs'])
Make predictions for custom files directly
pred = model.pipeline.predict(dl_kwargs, batch_size=4)
Get the model
library(reticulate)
kipoi <- import('kipoi')
model <- kipoi$get_model('labranchor')
Make a prediction for example files
predictions <- model$pipeline$predict_example()
Use dataloader and model separately
# Get the dataloader
setwd('~/.kipoi/models/labranchor')
dl <- model$default_dataloader(fasta_file='example_files/hg19.chr22.fa', gtf_file='example_files/hg19.chr22.gtf', length=70)
# get a batch iterator
it <- dl$batch_iter(batch_size=4)
# predict for a batch
batch <- iter_next(it)
model$predict_on_batch(batch$inputs)
Make predictions for custom files directly
pred <- model$pipeline$predict(dl_kwargs, batch_size=4)

Schema

Inputs

Dictionary of numpy arrays

Name: bidirectional_1_input

    Shape: (70, 4) 

    Doc: One-hot encoded RNA sequence


Targets

Single numpy array

Name: None

    Shape: (70,) 

    Doc: Predicted probability of being branchpoint along the sequence


Dataloader

Relative path: .

Version: 0.1

Doc: LaBranchoR predicts RNA splicing branchpoints using a Long Short-Term Memory network

Authors: Jun Cheng

Type: Dataset

License: MIT


Arguments

fasta_file : Reference genome sequence

gtf_file : file path; Genome annotation GTF file

length : length of considered candidate branchpoint region, upstream of 3'ss


Model dependencies
conda:

pip:
  • tensorflow>=1.0.0
  • keras>=2.0.4

Dataloader dependencies
conda:

pip:
  • pysam
  • numpy