A transcription factor is a regulatory protein that modulates the transcriptional velocity of genetic information from DNA to mRNA. This protein orchestrates the activation and deactivation of specific genes, thereby regulating their expression to coincide precisely with requisite phases and quantities throughout the cellular lifecycle. This paper introduces an innovative deep-learning methodology designed to ascertain the presence of transcription factors within protein sequences. This method, through a graph-based representation of the sequence, employs a Graph Neural Network for protein classification. We evaluated this novel approach across four datasets derived from the Swiss-Prot database; we achieved promising results that surpass those obtained through traditional fixed-length encodings and conventional machine-learning techniques and are comparable with state-of-the-art deep models.