Multi-spectral optoacoustic tomography (MSOT) combines rich contrast of optical imaging and high resolution of ultrasound, and becomes an attractive biomedical research tool in the last decade. Aligning MSOT images with anatomical map provided by magnetic resonance imaging (MRI) can potentially enhance the interpretation of optoacoustic signal which mainly reflects molecular and functional information. Therefore, developing an automated algorithm of image registration between MSOT and MRI is crucial. Existing MSOT-MRI registration algorithms mostly relied on manual segmentation, which requires user-dependent experience. Herein, we developed a fully automated algorithm for MSOT-MRI registration based on deep learning (DL). This workflow consists of DL-based segmentation and image transformation. We have experimentally demonstrated the accuracy and computational efficiency of the method, paving the way towards high-throughput MSOT data analysis in close future.