f5C-finder is the first neural network-based model for the identification on 5‑Formylcytidine modifications on mRNA.
This repository is the offical tensorflow vision of f5C-finder with all the codes used in the paper:
f5C-finder: combining language model with multi-head attention for predicting 5‑Formylcytidine modifications on mRNA
Your can use f5C-finder for inference at our webserver , submitted data will be kept confidential.
make sure your tensorflow > 2.0
You can download this Repository with
git clone https://github.com/TrumanPhd/f5C-finder.git
or just download with Github GUI
Ablation.py and Ablation_evaluation.py are the scripts for Ablation experiments
And for the Ablation results, please reference to the Ablation_exp.xlsx and Ablation_expRNN.xlsx
please reference: Model_lib2.py for LSTM-based models
Fusion3feature.py, encoder1234_best.py Hash1_best.py for attention-based models
(the five encoder can be referenced from feature_selection.py utils.py and Hao_keras_toolkit.py)
train.py and model_evaluation.py for model training and evaluation
the results can be referenced from metrics.xlsx
for RF ADA
can be referenced from ML.py ML_evaluation.py best_hyperparameters.xlsx
for SVM
can be referenced from SVM.py ML_evaluation.py SVMresults.xlsx
plot.py
and other files are the saved parameters from the model training.
In the dataset, we provide the data trained and evaluated from all the models, to make it more usable we replace 'U' to 'T',as shown in the dataset.
It`s my pleasure if can use f5C-finder for f5C identification or improve the accuarcy on the provided f5C dataset, and please cite at