RNA-Protein Interactions Prediction Using LSTM-CNN Technique

Abstract

An essential regulatory function of RNA-protein interaction is seen in human procedures. Investigating causes of illness and developing therapeutic targets can be aided by the evaluation and forecasting of RPIs. Additional study is being done to create computer techniques for RPI predictions that are less costly than the laborious and costly biological research investigations that were the mainstay of conventional RPI forecasting approaches. The current research paper suggests an approach for RPI forecasting that is understandable using LSTM-CNN. The current hybrid technique of LSTM-CNN is combined to efficiently analyze complex protein and RNA sequence information to pinpoint significant relationships among them. Due to this, we can forecast the interactions between RNA and proteins, which offers significant understanding regarding biological mechanisms and possible treatments. The success rate for forecasting obtained using the LSTM-CNN approach, determined by a five-fold cross-validation examination, is 96.3%, 87.3%, 91.7%, 92.9%, as well as 86.8% with samples RPI488, RPI369, RPI2241, RPI1807, along RPI1446. Using identical databases as cutting-edge RPI forecasting algorithms, our LSTM-CNN technique yields better precision. The accuracy of forecasting of the experiment information set surpasses that of various popular neural network computations, with five groups of evaluation sets made up of favorable specimens as well as five categories that include negative examples. The forecasting precision is 96.56%, 95.59%, 96.89%, 96.34%, as well as 95.87%, respectively. Furthermore, LSTM-CNN produced superior predictions across the datasets under examination. Numerous studies, including the ones presented here, demonstrate that LSTM-CNN is a trustworthy, precise, and consistent approach for RPI predictions.