Using deep learning methods to predict the VRC01 neutralization sensitivity by HIV-1 gp160 sequence features Open Access
Chu, Zhenghao (Spring 2020)
Introduction: The broadly neutralizing antibody (bnAb) VRC01 is being evaluated for its efficacy to prevent HIV-1 infection in the Antibody Mediated Prevention (AMP) trials. Our object is to applied Deep learning (DL) methods to see whether or not DL models can help improve accuracy on predicting sensitivity of neutralization of 611 HIV-1 Env pseudoviruses by VRC01.
Methods: We tried three different kinds of Deep Neural Network structures (FCNN, 1D-CNN, 1D-CNN+BiLSTM) to do the prediction and implemented a 5-fold cross-validation method to verify the performance of the model. We chose best model of each three neural network structures to do the prediction on our test set. We selected accuracy (Accuracy, Acc), precision (P), recall (R), F1 score (F1) and average area under the receiver operating characteristics (ROC) curve as evaluation indicators.
Results: The three mean AUCs (area under curve) for ROC curves are , , respectively. The prediction accuracies are 0.85, 0.83, 0.85 respectively.
Conclusion: For this small sample size task, our three Deep Learning models did not perform as well as the random forest model.
Key words: Human Immunodeficiency Virus (HIV), Antibody Mediated Prevention (AMP), Deep Learning, Deep Neural Network (DNN), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM)
Table of Contents
About this Master's Thesis
|Subfield / Discipline|
|Committee Chair / Thesis Advisor|
|Using deep learning methods to predict the VRC01 neutralization sensitivity by HIV-1 gp160 sequence features ()||2020-04-15 12:38:07 -0400||