NeoDiscovery/BassanilabNeoRanking

NeoRanking Immunogenic Neoantigen Prediction Model

Model Description

1.1 Base Model Implementation

This model is based on the methods described in the following publication:

Müller M, Huber F, Arnaud M, Kraemer A, Ricart Altimiras E, Michaux J, Taillandier-Coindard M, Chiffelle J, Murgues B, Gehret T, Auger A, Stevenson BJ, Coukos G, Harari A, Bassani-Sternberg M (2023)
Machine learning methods and harmonized datasets improve immunogenic neoantigen prediction
Immunity, 56, P2650-2663.E6, 10.1016/j.immuni.2023.09.002

• Paper Link: https://www.cell.com/immunity/fulltext/S1074-7613(23)00406-5
• Original Code Repository: https://github.com/bassanilab/NeoRanking

1.2 Feature Selection Differences

While the original paper by Müller et al. utilized the full feature set (available in supplementary materials), our implementation employs a refined subset of 8 key features:

• seq_len
• mutant_rank
• mutant_other_significant_alleles
• TAP_score
• mutant_rank_netMHCpan
• mut_Rank_Stab
• mutant_rank_PRIME
• mut_netchop_score_ct

1.3 Model Architecture

We exclusively use the XGBoost model for prediction and testing outputs.

Training Details

2.1 Training & Test Data

To minimize the effort of data preprocessing for reproducing the test, we adopted the data organization format from Neoranking, including file names. The training and test data are stored in data/Neopep_data_org.txt.

2.2 Training Parameters

Model training parameters follow the original code logic, with hyperparameter optimization performed using the hyperopt framework.

Model Usage

3.1 Setup and Installation

git clone https://github.com/bassanilab/NeoRanking
# Install and configure the required environment
cd NeoRanking
pip install -r requirements.txt

3.2 Configuration

Overwrite the NeoRanking/Utils/GlobalParameters.py file with our provided utils/GlobalParameters.py file. At the same time, we made modifications to the configure.sh in Neoranking, and the content is as follows.

# configure paths before running python scripts
export NEORANKING_RESOURCE="$pwd/Priorization/test"
export NEORANKING_CODE="$pwd"

mkdir -p "$NEORANKING_RESOURCE/data"
mkdir -p "$NEORANKING_RESOURCE/plots"
mkdir -p "$NEORANKING_RESOURCE/classifier_models"
mkdir -p "$NEORANKING_RESOURCE/classifier_results"
mkdir -p "$NEORANKING_RESOURCE/data/hla"
mkdir -p "$NEORANKING_RESOURCE/data/cat_encoding"

3.2 Model Parameters

Download the model parameters and place them in the Priorization/test/classifier_models directory.

3.3 Data Preparation

Place the test data file data/Neopep_data_org.txt into the Priorization/test/data path.

3.5 Execution

Run the test classifier from the NeoRanking/ directory:

bash test_classifer.sh

Special Notes

License Compliance

Our use of NeoRanking fully complies with the original bassanilab/NeoRanking license. If you use this model for citation or secondary applications, please adhere to the original author's citation rules and properly acknowledge the source.

Citation Requirements

When using this model, please cite the original work as specified in the NeoRanking repository and the associated publication.