UMAP-DBP: An Improved DNA-Binding Proteins Prediction Method Based on Uniform Manifold Approximation and Projection

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• DNA-binding proteins are crucial for cellular functions.
• Developing high-throughput methods for identifying DNA-binding proteins is essential.
• Machine learning and deep learning offer promising computational speed and accuracy.

Purpose of the Study:

• To propose a novel predictor, UMAP-DBP, for identifying DNA-binding proteins.
• To apply Uniform Manifold Approximation and Projection (UMAP) for feature selection in protein identification.
• To develop an efficient and accurate computational method for DNA-binding protein prediction.

Main Methods:

• Feature extraction from protein sequences using physicochemical distance transformation, Profile-based auto-cross covariance, and General series correlation pseudo amino acid composition.
• Feature selection employing Uniform Manifold Approximation and Projection (UMAP) and feature importance scores.
• Prediction of DNA-binding proteins using the Adaboost algorithm with a jackknife test validation.

Main Results:

• The UMAP-DBP predictor achieved an overall accuracy of 82.97% on the BP1075 dataset and 82.14% on the BP594 dataset.
• Cohen's kappa values of 0.66 and 0.64 were obtained for the BP1075 and BP594 datasets, respectively.
• Demonstrated the successful application of UMAP in conjunction with Adaboost for DNA-binding protein identification.

Conclusions:

• A feasible and accurate method (UMAP-DBP) has been developed for predicting DNA-binding proteins.
• This study represents the first successful application of UMAP for identifying DNA-binding proteins.
• The developed method and associated code are publicly available for further research and application.