Gene regulatory networks provide a powerful abstraction of the complex interactions among genes involved in functional pathways. Experimental determination of these interactions using a classical experimental method, although of extreme value, is laborious and prohibitive at large scales. Over the last decade, a number of computational approaches have been developed to infer gene regulatory networks from high-throughput experimental data. In this study, we introduce a new algorithm for regulatory network inference, based on stepwise multiple regression of time-series microarray data. Compared to other existing methods, our regression-based method provides a clear interpretation of the inferred interactions. The statistical significance associated with each prediction can be utilized to rank the interactions, which is important in prioritization of predictions for further experimental verification. We demonstrate the performance of our approach on a well-known yeast cell cycle pathway and show that it makes more accurate predictions than existing methods.