SaPt-CNN-LSTM-AR-EA: a hybrid ensemble learning framework for time series-based multivariate DNA sequence prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Biological sequence data mining is a critical area in bioinformatics.
• Biological sequences share similarities with time series data in representation and mechanism.
• Existing methods may not fully capture the complexities of biological sequence data.

Purpose of the Study:

• To propose a novel hybrid ensemble learning framework for biological time sequences (BTS).
• To represent biological sequences as time series for enhanced analysis.
• To improve the prediction performance and stability of biological sequence data mining.

Main Methods:

• Development of a hybrid ensemble learning framework named SaPt-CNN-LSTM-AR-EA.
• Construction of single-sequence and multi-sequence models using self-adaption pre-training one-dimensional convolutional recurrent neural network and autoregressive fractional integrated moving average fused evolutionary algorithm.
• Application of the framework to DNA sequence experiments involving six viruses.

Main Results:

• The SaPt-CNN-LSTM-AR-EA framework achieved good overall prediction performance.
• Prediction accuracy reached 1.7073, and correlation reached 0.9186 in DNA sequence experiments.
• The framework demonstrated superior effectiveness and stability compared to five benchmark models, increasing average accuracy by approximately 30%.

Conclusions:

• The proposed SaPt-CNN-LSTM-AR-EA framework offers significant advancements in biological sequence data mining.
• The approach of representing biological sequences as time series is effective for prediction.
• The framework has broad applicability in fields such as biology, biomedicine, computer science, sequence splicing, computational biology, and bioinformation.