Transcriptional regulation (TR) is a biological process in all living organisms by which cells respond to external stimuli, regulating the conversion of DNA to mRNA. The stimulus is driven by special transcription factor (TF) proteins. To control gene activities, TF can transit from inactive to active state, and bind to specific genes to activate or inhibit their transcription. This process finely tunes the amount of mRNA which is being produced through a variety of regulatory mechanisms, which is then translated into protein to act on the cellular environment. This makes key the understanding of regulatory mechanism for biological processes. However, while some gene-specific constants are relatively easy to quantify, experimental techniques to measure proteins concentrations, or to evaluate their effect on genes, are difficult and time consuming. In order to develop statistical approaches for transcription networks, statistical community has proposed several methods to infer activity levels of proteins, from time-series measurements of targets’ expression levels. Current simplified and approximated methods based on ordinary differential equation (ODE) are among the approaches most over-represented for modelling TR (e.g. single input motif, SIM, simplification). However, the implementation of SIM-based models for quick time-varying behaviour of protein stimulus may not be appropriate if TR models involve only a handful of genes and one TF. To deal with this drawback, a few number of approaches have been proposed in order to outperform the representation of fast switching time instants, but computational overheads are significant due to complex inference algorithms (e.g. MCMC methods). Using the theory related to latent force models (LFM), the development of this project provide a switched dynamical hybrid model based on Gaussian processes (GPs). To deal with discontinuities in the dynamical systems (or the latent driving force), an extension of the single input motif approach is introduced, that switches between different protein concentrations, and different dynamical systems. This creates a versatile representation for transcription networks that can capture discrete changes and non-linearities in the dynamics. The proposed method is evaluated on both simulated data and real data (e.g. Escherichia coli and Yeast datasets), concluding that our framework provides a computationally efficient statistical inference module of continuous-time concentration profiles, and allows an easy estimation of the associated model parameters.