Building models for gene regulation has been an important aim of Systems Biology over the past years, driven by the large amount of gene expression data that has become available. Models represent regulatory interactions between genes and transcription factors and can provide better understanding of biological processes, and means of simulating both natural and perturbed systems (e.g. those associated with disease). Gene regulatory network (GRN) quantitative modelling is still limited, however, due to data issues such as noise and restricted length of time series, typically used for GRN reverse engineering. These issues create an under-determination problem, with many models possibly fitting the data. However, large amounts of other types of biological data and knowledge are available, such as cross-platform measurements, knockout experiments, annotations, binding site affinities for transcription factors and so on. It has been postulated that integration of these can improve model quality obtained, by facilitating further filtering of possible models. However, integration is not straightforward, as the different types of data can provide contradictory information, and are intrinsically noisy, hence large scale integration has not been fully explored, to date. Here, we present an integrative parallel framework for GRN modelling, which employs evolutionary computation and different types of data to enhance model inference. Integration is performed at different levels. (i) An analysis of cross-platform integration of time series microarray data, discussing the effects on the resulting models and exploring crossplatform normalisation techniques, is presented. This shows that time-course data integration is possible, and results in models more robust to noise and parameter perturbation, as well as reduced noise over-fitting. (ii) Other types of measurements and knowledge, such as knock-out experiments, annotated transcription factors, binding site affinities and promoter sequences are integrated within the evolutionary framework to obtain more plausible GRN models. This is performed by customising initialisation, mutation and evaluation of candidate model solutions. The different data types are investigated and both qualitative and quantitative improvements are obtained. Results suggest that caution is needed in order to obtain improved models from combined data, and the case study presented here provides an example of how this can be achieved. Furthermore, (iii), RNA-seq data is studied in comparison to microarray experiments, to identify overlapping features and possibilities of integration within the framework. The extension of the framework to this data type is straightforward and qualitative improvements are obtained when combining predicted interactions from single-channel and RNA-seq datasets.