VEGF-ERCNN: A deep learning-based model for prediction of vascular endothelial growth factor using ensemble residual CNN

Abstract

Vascular Endothelial Growth Factor (VEGF), a signaling protein family, is essential in angiogenesis, regulating the growth and survival of endothelial cells that create blood vessels. VEGF is critical in osteogenesis for coordinating blood vessel growth with bone formation, resulting in a well-vascularized environment that promotes nutrition and oxygen delivery to bone-forming cells. Predicting VEGF is crucial, yet experimental methods for identification are both costly and time-consuming. This paper introduces VEGF-ERCNN, an innovative computational model for VEGF prediction using deep learning. Two datasets were generated using primary sequences, and a novel feature descriptor called multi fragmented-position specific scoring matrix-discrete wavelet transformation (MF-PSSM-DWT) was developed to extract numerical characteristics from these sequences. Model training is performed via deep learning techniques such as generative adversarial network (GAN), gated recurrent unit (GRU), ensemble residual convolutional neural network (ERCNN), and convolutional neural network (CNN). The VEGF-ERCNN outperformed other competitive predictors on both training and testing datasets by securing the highest 92.12 % and 83.45 % accuracies, respectively. Accurate prediction of VEGF therapeutic targeting has transformed treatment techniques, establishing it as a crucial participant in both health and disease.