New Features Developed for the Detection of a Promoter Based on Machine Learning

A promoter is a brief section of DNA that initiates RNA polymerase transcription of a gene. Usually, it's situated right upstream of the transcription start point. Numerous human illnesses, including diabetes, cancer, and Huntington's disease, have been shown to have a genetic promoter as their major cause. As a result, categorizing promoters has emerged as a fascinating issue that many bioinformatics experts are interested in studying. Numerous studies were carried out to address this issue. However, their performance outcomes still need to be improved. This study presents a challenge by providing a novel set of engineered capabilities to improve classification between promoter and non-promoter sequence (PS and n_PS) types. The urgency of the studies lies inside the want for more interpretable, green models that may work nicely with restricted biological statistics. have been compared between the extracted classical and newly developed features (such as; consisting of nucleotide composition and counting, G_C content material, k-mer frequency, and sequence complexity) classify to differentiate PS from nPS. Using the MATLAB environment, we evaluated five device learning models (SVM, KNN, LR, DT, and NB) at the benchmark UCI dataset. The proposed feature set significantly improved the category's overall performance. Naïve Bayes achieved a percentage of 100 % for accuracy, sensitivity, specificity and F1score when using the new feature structure. In contrast, accuracy, sensitivity and F1-score were 94%, 93% and 90% respectively when basic features were used. These consequences validate the effectiveness of the developed capabilities and their biological interpretability. This looks at a promising technique for promoter prediction with interpretable, reproducible consequences. Future work will consist of validation on large datasets, including EPD and ENCODE, to ensure broader applicability.