Classical machine learning models to predict TE

Predicting the translation efficiency of messenger RNA in mammalian cells
Dinghai Zheng, Logan Persyn, Jun Wang, Yue Liu, Fernando Ulloa Montoya, Can Cenik, Vikram Agarwal
bioRxiv 2024.08.11.607362; doi: https://doi.org/10.1101/2024.08.11.607362

Environemnt Setup

1. Create conda environment:
   conda env create -f environment.yml --prefix ./TE_classic_ML_env/
2. Activate conda environment:
   conda activate ./TE_clasic_ML_env

Data

Training data is already provided in ./data/.

If generating training data yourself:

• Place/symlink appris_human_v2_selected.fa and appris_mouse_v2_selected.fa in ./data/.
• Place/symlink human_all_biochem_feature_no_len.csv in ./biochem_and_struct_data if training with biochem data.

Training

Training examples can be found in experiments.py. Use -e human_all_no_struct argument to train all human models (save models with -s flag).

Predicting

Predict with predict.py. Example inputs and outputs can be found in ./examples. Full command example:
python .\predict.py --model_dir ./results/human/all_cell_lines/lgbm-LL_P5_P3_CF_AAF_3mer_freq_5/ --data_path ./examples/predict_input_example.csv --output_path ./examples/predict_output_example.csv