firdaaminy/cerebro-paper

Cerebro

Cerebro is a machine-learning classifier for somatic mutation detection based on an extremely randomized trees model.

This package includes two scripts for training and executing the Cerebro model for somatic mutation identification from whole- exome next-generation sequencing data. Required input is a set of candidate variants and informative features as described below.

Requirements

This code requires the following dependencies

• Python (v2.7) https://www.python.org/
• numpy (v1.8) http://www.numpy.org/
• scikit-learn (0.19.1) http://scikit-learn.org/stable/index.html

Training and Testing Datasets

Training data must consist of tab-delimited integers or decimal values, with the exception of the first two columns which are formatted as:

• a unique identifier per mutation candidate
• the mutation type (SBS/INS/DEL)
• a final column consisting of either 0 (for known incorrect calls) or 1 for true somatic mutations

See the file example_points.txt for an example.

Testing data follows the same format as the training data, but without the final column of 0s and 1s.

---

Usage

To train the model, execute the command:

mkdir trained_model && ./cerebro_train.py example_points.txt trained_model

To run the trained model on new data, execute the command

./cerebro_call.py testing_points.txt trained_model > classifications.txt

README documentation

2017-12-15 -- Source code and example training data for Cerebro somatic mutation confidence scoring.

This package includes two scripts for training and executing the Cerebro model for somatic mutation identification from whole- exome next-generation sequencing data. Required input is a set of candidate variants and informative features.

Training Data Example (example_points.txt) This data must consist of tab-delimited integers or decimal values, with the exception of the first two columns which are formatted as:

1. a unique identifier per mutation candidate
2. the mutation type (SBS/INS/DEL)

The final column must consist of either 0 (for incorrect calls) or 1 for true somatic mutations.

To train the model, execute the command: mkdir trained_model && ./cerebro_train.py example_points.txt trained_model

To run the trained model on new data, execute the command ./cerebro_call.py testing_points.txt trained_model