Jian Gao

Predicting the biological properties of molecules is crucial in computer-aided drug development, yet it's often impeded by data scarcity and imbalance in many practical applications. Existing approaches are based on self-supervised learning or 3D data and using an increasing number of parameters to improve performance. These approaches may not take full advantage of established chemical knowledge and could inadvertently introduce noise into the respective model. In this study, we introduce a more elegant transformer-based framework with focused attention for molecular representation (TransFoxMol) to improve the understanding of artificial intelligence (AI) of molecular structure property relationships. TransFoxMol incorporates a multi-scale 2D molecular environment into a graph neural network + Transformer module and uses prior chemical maps to obtain a more focused attention landscape compared to that obtained using existing approaches. Experimental results show that TransFoxMol achieves state-of-the-art performance on MoleculeNet benchmarks and surpasses the performance of baselines that use self-supervised learning or geometry-enhanced strategies on small-scale datasets. Subsequent analyses indicate that TransFoxMol's predictions are highly interpretable and the clever use of chemical knowledge enables AI to perceive molecules in a simple but rational way, enhancing performance.

Ovarian cancer is currently the second most common malignant tumor among gynecological cancers worldwide, primarily due to challenges in early diagnosis, high recurrence rates, and resistance to existing treatments. Current therapeutic options are inadequate for addressing the needs of ovarian cancer patients. Ferroptosis, a novel form of regulated cell death with demonstrated tumor-suppressive properties, has gained increasing attention in ovarian malignancy research. A growing body of evidence suggests that ferroptosis plays a significant role in the onset, progression, and incidence of ovarian cancer. Additionally, it has been found that immunotherapy, an emerging frontier in tumor treatment, synergizes with ferroptosis in the context of ovarian cancer. Consequently, ferroptosis is likely to become a critical target in the treatment of ovarian cancer.

Children with untreated coeliac disease (CD) may develop enamel defects. Moreover, children with untreated CD have increased serum levels of gliadin reactive IgG, which may cross-react to amelogenin. The aim of this study was to investigate reactivity of anti-gliadin IgA and IgG to amelogenin in children with untreated CD. Blood samples from patients with CD (n = 75) and from disease controls (n = 24) were analysed for IgA and IgG reactivities to amelogenin (Emdogain) and to gliadin by ELISA. Whereas children with CD had statistically significantly higher serum levels of anti-amelogenin IgA, only those with the most severe CD (Marsh 3c) had significantly higher anti-amelogenin IgG immune reactivity than the disease controls. Western blotting confirmed that the IgA and IgG immune reactivity was to the amelogenin-specific bands in Emdogain and to a 22-kDa human recombinant amelogenin. Cross-inhibition studies revealed that the anti-amelogenin immune reactivity was not only caused by anti-gliadin cross-reactivity but also included amelogenin selective immune reactivity. Some controls had high levels of anti-amelogenin IgA and IgG, similar to children with CD. Thus, anti-amelogenin IgA and IgG may not only be involved in the aetiology of CD-associated enamel defects but may also interfere with enamel maturation in non-coeliac children.

Breast cancer (BC) is the most frequently diagnosed malignancy among women worldwide, with a high incidence and mortality rate. Despite advances in treatment, approximately 10%–15% of patients with BC still face recurrence. Therefore, improving BC therapy remains a significant challenge. In this article, we provide a detailed overview, categorizing and elaborating the developments of current research progress on nanodrug delivery systems based on ferroptosis for BC treatment. By increasing the iron content in BC cells and inhibiting the defense system against ferroptosis, the accumulation of lipid peroxides is promoted, and ferroptosis is induced in BC cells. In addition to directly targeting tumor cells, nanodrug delivery systems can remodel the tumor microenvironment, inhibit BC primary growth, and prevent distant metastasis. These nanomaterials, after drug loading and modification, possess characteristics such as smart activation, controlled release, specific targeting, good biocompatibility, and long circulation time, thereby enhancing the efficacy of BC treatment. We also classify and discuss the mechanisms and advantages of different types of nanomaterials. Finally, we discuss how multifunctional nanosystems can sensitize ferroptosis when combined with radiotherapy, chemotherapy, immunotherapy, and phototherapy to achieve synergistic effects in BC treatment. This work reveals the potential of ferroptosis-based nanomaterials in overcoming BC, analyzes the limitations of the clinical application and proposes possible solutions, offering a promising direction for future treatment strategies. 1. Ferroptosis-based nanodrugs induce ferroptosis by increasing intracellular iron levels, disrupting redox homeostasis, and promoting LPO accumulation. 2. Nanodrugs not only target BC cells, but also target TME. 3. Ferroptosis-based nanodrugs can also synergize with chemotherapy, radiotherapy, immunotherapy, PDT, PTT, and other treatment modalities for enhanced BC management.

The compute-in-memory (CIM) paradigm holds great promise to efficiently accelerate machine learning workloads. Among memory devices, static random-access memory (SRAM) stands out as a practical choice for its exceptional reliability in the digital domain and excellent scalability. Recently, there has been a growing interest in accelerating floating-point (FP) deep neural networks (DNNs) with SRAM CIM due to their critical importance in DNN training and high-accurate inference. This paper proposes an energy-efficient SRAM CIM macro for FP DNNs. To achieve the design, we identify a lightweight approach that decomposes conventional FP mantissa multiplication into two parts: mantissa sub-addition (sub-ADD) and mantissa sub-multiplication (sub-MUL). Our study shows that while mantissa sub-MUL is compute-intensive, it only contributes to the minority of FP products, whereas mantissa sub-ADD, although compute-light, accounts for the majority of FP products. Recognizing "Addition is Most You Need", we develop a novel hybrid-domain SRAM CIM macro to accurately handle mantissa sub-ADD in the digital domain while improving the energy efficiency of mantissa sub-MUL using analog computing. Experiments with the MLPerf benchmark show its remarkable improvement in energy efficiency on average by 3×~ 3.6× (2.5×~3.1×) in inference (training) compared to a fully digital baseline without any accuracy loss, showcasing its great potential for FP DNN acceleration.