Category

Archives

Overfit deep neural network for predicting drug-target interactions

Drug-target interactions (DTIs) prediction is an important step in drug discovery. As traditional biological experiments or high-throughput screening are high cost and time-consuming, many deep learning models have been developed. Overfitting must be avoided when training deep learning models. We propose a simple framework, called OverfitDTI, for DTI prediction. In OverfitDTI, a deep neural network (DNN) model is overfit to sufficiently learn the features of the chemical space of drugs and the biological space of targets. The weights of trained DNN model form an implicit representation of the nonlinear relationship between drugs and targets. Performance of OverfitDTI on three public datasets showed that the overfit DNN models fit the nonlinear relationship with high accuracy. We identified fifteen compounds that interacted with TEK, a receptor tyrosine kinase contributing to vascular homeostasis, and the predicted AT9283 and dorsomorphin were experimentally demonstrated as inhibitors of TEK in human umbilical vein endothelial cells (HUVECs).

 

Comments:

It sounds like you've described a novel approach, OverfitDTI, for predicting drug-target interactions (DTIs) using deep neural networks (DNNs). The key aspects of your approach involve addressing the challenges associated with traditional experimental methods and high-throughput screening, such as cost and time constraints, by employing machine learning techniques. Here's a summary of the main points from your description:

1. **Challenge in DTI Prediction:** Predicting drug-target interactions is crucial in drug discovery, but traditional methods like biological experiments and high-throughput screening are expensive and time-consuming.

2. **Use of Deep Learning Models:** To address these challenges, deep learning models have been employed due to their ability to learn complex patterns from large datasets.

3. **OverfitDTI Framework:** You propose a framework called OverfitDTI, where a DNN model is intentionally overfit to learn the intricate features of the chemical space of drugs and the biological space of targets. By overfitting, the model captures the nuanced relationships between drugs and targets, which might be missed in simpler models.

4. **Implicit Representation:** The weights of the trained DNN model serve as an implicit representation of the nonlinear relationship between drugs and targets. This representation encapsulates the learned knowledge about DTIs.

5. **Performance Evaluation:** OverfitDTI was evaluated on three public datasets. The results demonstrate that the overfitted DNN models accurately capture the nonlinear relationships between drugs and targets, achieving high prediction accuracy.

6. **Case Study:** As a case study, OverfitDTI successfully identified fifteen compounds that interact with TEK, a receptor tyrosine kinase involved in vascular homeostasis. Two of the predicted compounds, AT9283 and dorsomorphin, were experimentally confirmed as inhibitors of TEK in human umbilical vein endothelial cells (HUVECs).

This approach appears to be a creative solution to the challenge of predicting DTIs. By intentionally overfitting the model, you allow it to grasp intricate patterns within the data, leading to accurate predictions. The experimental validation of the predictions further strengthens the credibility of your method.

Related Products

Cat.No. Product Name Information
S1134 AT9283 AT9283 is a potent JAK2/3 inhibitor with IC50 of 1.2 nM/1.1 nM in cell-free assays; also potent to Aurora A/B, Abl1(T315I).

Related Targets

Aurora Kinase Bcr-Abl JAK