Jul 2025· Chemical Science· Vol 16, pp. 14698-14709· 4 citations· 67 references
Medicine
TL;DR
This work proposes a novel strategy that incorporates two models: a fragment-based model (Frag-MD) and a character-based model (MD), both integrating a conditional transformer, recurrent neural networks, and reinforcement learning that holds the potential to establish a new paradigm for discovering novel PSs applicable in PDT.
Abstract
Photodynamic therapy (PDT) is a clinically approved therapeutic modality that has demonstrated significant potential for cancer treatment, and triplet photosensitizers (PSs) play a key role in its efficacy. Despite deep learning having emerged as a next-generation tool for material discovery, existing methods mainly target a limited subset of triplet PSs, such as thermally activated delayed fluorescence (TADF) materials, neglecting the critical intersystem crossing (ISC) between the high-lying singlet and triplet states (ΔESnTn). To overcome this limitation, we compiled a comprehensive dataset (∼1.90 × 109) of triplet PSs encompassing various ISC mechanisms. Then, we proposed a novel strategy that incorporates two models: a fragment-based model (Frag-MD) and a character-based model (MD), both integrating a conditional transformer, recurrent neural networks, and reinforcement learning. In silico experiments revealed that the Frag-MD model outperforms the MD model in generating larger conjugated motifs with higher average ring numbers and atom counts; while the MD model generates twice as many unique motifs and excels in novelty and diversity, as evaluated by conditional and MOSES metrics. Therefore, our approach is highly effective for modifying conjugated motifs and designing novel triplet PSs. Notably, the recently reported high-efficiency triplet PSs have been re-identified through ablation experiments using our proposed models, which target ΔESnTn and significantly outperform traditional baselines, achieving a prediction accuracy of 73% versus 4%. Our approach holds the potential to establish a new paradigm for discovering novel PSs applicable in PDT.
In recent years, heavy-atom-free photosensitizers have been recognized as an important class of agents for photodynamic therapy (PDT). In particular, redox-activated heavy-atom-free photosensitizers inspired by the redox imbalance of the tumor microenvironment are emerging as a promising strategy to improve the specificity and therapeutic efficiency of PDT. This review highlights design strategies of redox-activated heavy-atom-free photosensitizers (PSs), including donor-acceptor-based frameworks, thionation, aggregation-induced emission (AIE) driven intersystem crossing enhancement, and the integration of responsive moieties for reactive oxygen species (ROS), glutathione (GSH), cysteine (Cys), and hydrogen sulfide (H₂S). In addition, recent advances in the development of redox-activated heavy-atom-free PSs over the past three years are summarized. Finally, main challenges, including hypoxic tumors, limited tumor-targeting efficiency, and limited light penetration into deep-seated tumors, as well as future prospects for this field, are discussed.
Hyunsun Jeong, Hao-Yang Song, V. Nguyen et al.· Advanced Drug Delivery Revie...· 1 citation
Photodynamic therapy (PDT) is a minimally invasive therapeutic modality based on the interaction between a photosensitizer (PS), light, and molecular oxygen to generate reactive oxygen species (ROS) capable of inducing localized cytotoxicity. Natural products provide a chemically diverse source of photosensitizers, including curcumin, hypericin, hypocrellin, chlorin derivatives, alkaloids, flavonoids, anthraquinones, and other photoactive scaffolds. However, their translational development remains limited by poor solubility, aggregation, instability, variable purity, limited tissue penetration, suboptimal pharmacokinetics, and insufficient formulation readiness. In parallel, artificial intelligence (AI), including machine learning (ML), deep learning (DL), quantitative structure–activity relationship (QSAR) and quantitative structure–property relationship (QSPR) modeling, radiomics, and predictive analytics, is increasingly being applied to photosensitizer discovery, molecular property prediction, nanoformulation optimization, treatment planning, and precision PDT. This critical review evaluates the intersection between AI, natural photosensitizers, nanopharmaceutical development, and PDT, with emphasis on methodological strengths, current limitations, and translational priorities. A PRISMA 2020-inspired search strategy identified 27 studies for qualitative synthesis, comprising 11 review articles and 16 original investigations, while additional seminal references were used for historical and mechanistic contextualization. The analysis indicates that current AI applications in PDT are concentrated around molecular property prediction, QSAR/QSPR modeling, phototoxicity assessment, radiomics, image-guided therapy, and treatment-response prediction, whereas AI-guided exploration of natural photosensitizer chemical space and AI-assisted nanoformulation design remain comparatively underdeveloped. Key barriers include heterogeneous datasets, limited natural-product representation in predictive models, insufficient external validation, weak integration between formulation variables and photodynamic outcomes, and limited consideration of manufacturing and regulatory requirements. This review proposes an integrated AI-enabled translational framework connecting natural-product chemical diversity, photochemical prediction, nanocarrier optimization, precision PDT validation, and clinical implementation.
R. Gonçalves, E. Costa· Pharmaceutics· 0 citations
Cancer remains a major threat to global public health, highlighting the urgent need for precise and effective treatment strategies. Photothermal therapy (PTT) has gained increasing attention as a promising alternative owing to its non‐invasiveness, deep tissue penetration, and oxygen‐independent mechanism. However, the development of high‐performance organic photothermal agents (PTAs) with pronounced near‐infrared absorption, high photothermal conversion efficiency (PCE), and good photostability remains challenging. Herein, we propose a synergistic strategy combining π‐bridge and donor engineering to design a series of aggregation‐induced emission (AIE)‐active small‐molecule photosensitizers. Experimental and theoretical analyses reveal that stepwise introduction of a planar thiophene π‐bridge and methoxy donor groups systematically regulates the excited‐state energy dissipation pathways, enabling a continuous modulation from photodynamic therapy (IsoTPA), through balanced synergistic PDT/PTT (IsoTHTPA, PCE = 57.2%), to highly efficient PTT (IsoTHTO, PCE = 64.3%), which reveals the fundamental competitive interplay between intersystem crossing and non‐radiative decay. Encapsulated into nanoparticles, IsoTHTO NPs demonstrate efficient photothermal ablation of 4T1 tumor cells and significant tumor growth inhibition under 660 nm laser irradiation both in vitro and in vivo. This work presents a reasonable molecular design approach for tunable access to PDT, synergistic PDT/PTT, and PTT modalities from a single molecular platform for cancer phototheranostics.
Yin Li, Dong He, Lu Liu et al.· Advanced Healthcare Material...· 0 citations
Laryngeal cancer requires precise and minimally invasive therapeutic strategies to preserve critical physiological functions. Herein, an oxygen-embedded fused-ring organic semiconductor (COi8FIC) was reported, which is formulated into multifunctional nanoparticles for near-infrared (NIR)-triggered multimodal phototherapy. The incorporation of oxygen atoms into the conjugated backbone, together with fluorinated cyanoindanone end-groups, narrows the bandgap to 1.35 eV and enhances intramolecular charge transfer, thereby promoting non-radiative decay and intersystem crossing. As a result, COi8FIC nanoparticles exhibit a high photothermal conversion efficiency of 62% and generate multiple reactive oxygen species (•OH, O2-, and 1O2) via synergistic Type I and Type II photodynamic pathways. In vitro, the nanoparticles produce approximately 2.5-fold higher ROS levels than monotherapies. In vivo, NIR irradiation induces rapid hyperthermia (ΔT ca. 23.6 °C), resulting in a tumor growth inhibition rate of 93.3%. Tumor regression was observed in all treated mice, with complete tumor eradication in three of five mice and near-complete elimination in the remaining two, with no significant irreversible organ damage. These results highlight the potential of this system as a promising platform for multimodal phototherapy of laryngeal cancer.
Hao Liu, Bo Yu, Shiwen Zhong et al.· ACS Applied Materials and In...· 0 citations
A new method for surgically removing training examples from a model reveals that as datasets grow, the link between what a model learns and what it produces dissolves.
What does it take to trust AI-driven HVAC optimization? Our AI Model Factory combines agents, machine learning, reinforcement learning and deterministic checks in a governed workflow designed for messy, real-world building data. The post We built an AI factory for HVAC control appeared first on GPT-Lab.
Microsoft Research Blog· microsoft.comJul 30, 2026
LLMs do not get smarter just by remembering more. EvoLib turns experience into evolving knowledge, taking reusable skills and insights that help models learn and adapt across tasks long after deployment. The post EvoLib: Turning experience into evolving knowledge appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduMay 20, 2026
