Cancer immunotherapy has attracted great attention as a potential therapeutic approach for advanced malignancies due to its promising survival benefits. Comprehension of intricate interactions between the tumor microenvironment (TME) and immune checkpoint inhibitors (ICIs) is crucial for optimizing and improving immunotherapies. Currently, several experimental strategies are available to monitor this complexity but most of them fail to facilitate real-time monitoring of the immune response such as cellular phagocytosis and cytolysis. Consequently, the application of intravital imaging has been extensively studied in the domain of cancer immunotherapy. Intravital imaging has been proven to be a powerful real-time imaging modality that provides insights into intratumoral immune responses, cellular metabolic signatures, tumor vasculature, and cellular functions. This review aims to provide a comprehensive overview of the latest research on intravital imaging in cancer immunotherapy, especially addressing how intravital imaging sheds light on essential features of tumor immunity, immune infiltrations, tumor angiogenesis, and aids in the clarification of underlying immunotherapeutic mechanisms. Moreover, a variety of labeling tools, imaging windows and models for real-time visualizations of TME are also summarized. We will also investigate the full potential of using intravital imaging to circumvent the limitations of currently available imaging modalities, which hold promise to advent efficient immunotherapy for cancer patients. Graphical abstract

Background: Cancer immunotherapy has revolutionized the clinical management of cancer due to its promising survival benefits. Although FDA-approved cancer immunotherapy primarily leverages adaptive immunity for therapeutic efficacy, the critical role of innate immunity in tumor surveillance and eradication has been increasingly recognized. Recently, with the deepening of research in this field, macrophages have emerged as key effectors of the innate anti-tumor response, with antibody-dependent cellular phagocytosis (ADCP) becoming one of the primary mechanisms involved in cancer immunotherapy. As such, exploring ADCP-inducing monoclonal antibodies (mAbs) as forefront therapeutic regimens for patients with cancer is therefore highly sought after. Main body: Here, we broadly summarize recent translational advances in harnessing macrophage phagocytosis as a pivotal therapeutic approach in cancer immunotherapy, especially addressing how intravital imaging sheds light on ADCP in a real-time manner. Moreover, various challenges for targeting phagocytosis, strategies to enhance ADCP, and factors limiting its efficiency have also been summarized. We also investigated the potential of chimeric antigen receptor (CAR) macrophages as a next-generation therapeutic modality that bridges the innate and adaptive immune systems to induce robust anti-tumor immune responses. Conclusions: The insights presented in this review will set the scene for future investigations of possible alternative approaches that consider the antibody-dependent cellular phagocytosis for improved cancer immunotherapy.

AI models for drug discovery and chemical literature mining must interpret molecular images and generate outputs consistent with 3D geometry and stereochemistry. Most molecular language models rely on strings or graphs, while vision-language models often miss stereochemical details and struggle to map continuous 3D structures into discrete tokens. We propose DeepMoLM: Deep Molecular Language M odeling, a dual-view framework that grounds high-resolution molecular images in geometric invariants derived from molecular conformations. DeepMoLM preserves high-frequency evidence from 1024 $\times$ 1024 inputs, encodes conformer neighborhoods as discrete Extended 3-Dimensional Fingerprints, and fuses visual and geometric streams with cross-attention, enabling physically grounded generation without atom coordinates. DeepMoLM improves PubChem captioning with a 12.3% relative METEOR gain over the strongest generalist baseline while staying competitive with specialist methods. It produces valid numeric outputs for all property queries and attains MAE 13.64 g/mol on Molecular Weight and 37.89 on Complexity in the specialist setting. On ChEBI-20 description generation from images, it exceeds generalist baselines and matches state-of-the-art vision-language models. Code is available at https://github.com/1anj/DeepMoLM.