Unveiling the genetic symphony: Deep learning for decoding promoters and non-promoters in DNA sequence

Abstract

Promoters are a significant area in the structure of DNA which is vital for the transcription of the exact gene in the genome. There are several types of promoters in the DNA that performs certain functions. The Mutations in the promoters are the major reason for many diseases like cancer, diabetes, etc. Effective identification of promoters is important for the diagnosis of certain diseases. The traditional identification of promoters is expensive and time-consuming. To resolve the issue, several conventional methods attempted to achieve better promoter and non-promoter identification systems but lack limitations like handling larger datasets, speed, and accuracy. Therefore, the projected system employs Bi-GRU (Bi-Gated Recurrent Units) with M-AM (Modified Attention Mechanism) to detect promoters and non-promoters in DNA sequence. Bi-GRU is utilized in the projected system for less complexity performance, better convergence speed, and the ability to work fast with huge data. Though it is widely utilized for detecting promoters and non-promoters, it has minor limitations, like handling larger datasets. To resolve this, the projected system utilizes Bi-GRU with M-AM to focus on the significant features of the data, which enhances the model's efficacy. M-AM of the numerical significances measure and sequence context weights. Promoters and non-promoters in the DNA sequence dataset are utilized in the projected system. Formerly, the performance of the projected system was calculated with evaluation metrics. Further, the proposed detection method is compared internally with LSTM (Long short-term Memory), BI-LSTM, and GRU to evaluate the model's efficiency. The projected detection of promoters and non-promoters is intended to assist researchers and physicians in genomics and molecular biology to enhance the diagnosis of the disease and disorders caused by the mutations of promoters.