Gene Regulatory Networks are major mechanisms governing biological processes. These networks are often inferred from gene expression matrices using Machine Learning algorithms. Nevertheless this task requires large datasets, to avoid learning unsuitable links due to over-fitting. This limitation impedes applying safely such tools on gene expression datasets with limited number of samples, such as rare cell-types or diseases. Moreover, the intrinsic black-box nature of most Machine Learning tools is an important drawback in sensitive areas such as in the bio-medical context. In this work we propose a new method, based on Transfer Learning and Semi-Supervised Learning, to infer interpretable gene regulatory networks, when only few samples are available. We have applied this tool to infer the first gene regulatory network of cell-line SH-SY5Y, a well recognized in vitro model for Alzheimer’s disease. This biomedical application is particularly important, since Alzheimer’s disease is a poorly understood neurodegenerative disease that causes around 60–70% of dementia cases worldwide and since there are currently no treatments to stop or reverse it. Our transfer methodology was successfully assessed qualitatively, and we revealed that our simple and interpretable models remained competitive w.r.t state-of-the-art complex models. The exploration of the regulatory links, revealed that our methodology identified well-known regulatory pathways involved in Alzheimer’s disease. The encouraging outcomes yielded by our approach suggest its potential for analyzing other small gene expression datasets.