# iPro-CSAF: identification of promoters based on convolutional spiking neural networks and spiking attention mechanism

## TL;DR

This paper introduces iPro-CSAF, a new computational method using spiking neural networks to accurately identify DNA promoters across multiple species.

## Contribution

The novel contribution is the development of iPro-CSAF, a convolutional spiking neural network with spiking attention mechanism for promoter recognition.

## Key findings

- iPro-CSAF outperforms existing methods in promoter recognition with fewer network parameters.
- The method achieves good performance on both prokaryotic and eukaryotic promoters from seven species.
- iPro-CSAF demonstrates low complexity and strong generalization for promoter identification.

## Abstract

A promoter is a DNA segment which plays a key role in regulating gene expression. Accurate identification of promoters is significant for understanding the regulatory mechanisms involved in gene expression and genetic disease treatment. Therefore, it is an urgent challenge to develop computational methods for identifying promoters. Most current methods were designed for promoter recognition on few species and required complex feature extraction methods in order to attain high recognition accuracy. Spiking neural networks have inherent recurrence and use spike-based sparse coding. Therefore, they have good property of processing spatio-temporal information and are well suited for learning sequence information. In this study, iPro-CSAF, a convolutional spiking neural network combined with spiking attention mechanism is designed for promoter recognition. The method extracts promoter features by two parallel branches including spiking attention mechanism and a convolutional spiking layer. The promoter recognition of iPro-CSAF is evaluated by exhaustive promoter recognition experiments including both prokaryotic and eukaryotic promoter recognition from seven species. Our results show that iPro-CSAF outperforms promoter recognition methods which used parallel CNN layers, methods which combined CNNs with capsule networks, attention mechanism, LSTM or BiLSTM, and CNNs-based methods which needed priori biological or text feature extraction, while our method has much fewer network parameters. It indicates that iPro-CSAF is an effective computational method with low complexity and good generalization for promoter recognition.