Cheng Ge

Conotoxins are small, disulfide-rich peptides that display exceptional affinity and selectivity for ion channels and receptors, making them valuable templates for therapeutic development. However, their optimization remains challenging due to the limited diversity of naturally occurring variants and the labor-intensive nature of conventional engineering strategies. Here, we present CreoPep, a deep learning-based generative framework specifically developed to design and optimize conotoxins targeting defined receptors. CreoPep integrates masked language modeling with a progressive masking scheme and employs an augmentation pipeline that combines physics-based energy screening with temperature-controlled multinomial sampling. This enables the generation of structurally and functionally diverse peptide variants while retaining essential pharmacological features. Structural analysis shows that CreoPep-generated variants adopt both conserved and previously unobserved binding modes, including disulfide-deficient forms. Together, these findings establish CreoPep as a powerful computational-experimental framework for the rational design of conotoxin-based peptides and provide a foundation for extending similar approaches to other peptide families.

Conotoxins are a class of peptide toxins secreted by marine mollusks of the Conus genus, characterized by their unique mechanism of action and significant biological activity, making them highly valuable for drug development. However, traditional methods of acquiring conotoxins, such as in vivo extraction or chemical synthesis, face challenges of high costs, long cycles, and limited exploration of sequence diversity. To address these issues, we propose the ConoGPT model, a conotoxin sequence generation model that fine-tunes the ProtGPT2 model by incorporating disulfide bond information. Experimental results demonstrate that sequences generated by ConoGPT exhibit high consistency with authentic conotoxins in physicochemical properties and show considerable potential for generating novel conotoxins. Furthermore, compared to models without disulfide bond information, ConoGPT outperforms in terms of generating sequences with ordered structures. The majority of the filtered sequences were shown to possess significant binding affinities to nicotinic acetylcholine receptor (nAChR) targets based on molecular docking. Molecular dynamics simulations of the selected sequences further confirmed the dynamic stability of the generated sequences in complex with their respective targets. This study not only provides a new technological approach for conotoxin design but also offers a novel strategy for generating functional peptides.

With the emergence of multidrug-resistant bacteria, antimicrobial peptides (AMPs) offer promising options for replacing traditional antibiotics to treat bacterial infections, but discovering and designing AMPs using traditional methods is a time-consuming and costly process. Deep learning has been applied to the de novo design of AMPs and address AMP classification with high efficiency. In this study, several natural language processing models were combined to design and identify AMPs, i.e. sequence generative adversarial nets, bidirectional encoder representations from transformers and multilayer perceptron. Then, six candidate AMPs were screened by AlphaFold2 structure prediction and molecular dynamic simulations. These peptides show low homology with known AMPs and belong to a novel class of AMPs. After initial bioactivity testing, one of the peptides, A-222, showed inhibition against gram-positive and gram-negative bacteria. The structural analysis of this novel peptide A-222 obtained by nuclear magnetic resonance confirmed the presence of an alpha-helix, which was consistent with the results predicted by AlphaFold2. We then performed a structure–activity relationship study to design a new series of peptide analogs and found that the activities of these analogs could be increased by 4–8-fold against Stenotrophomonas maltophilia WH 006 and Pseudomonas aeruginosa PAO1. Overall, deep learning shows great potential in accelerating the discovery of novel AMPs and holds promise as an important tool for developing novel AMPs.

Nasopharyngeal carcinoma (NPC) is one of the most common head and neck cancers. Early diagnosis plays a critical role in the treatment of NPC. To aid diagnosis, deep learning methods can provide interpretable clues for identifying NPC from magnetic resonance images (MRI). To identify the optimal models, we compared the discrimination performance of hierarchical and simple layered convolutional neural networks (CNN). Retrospectively, we collected the MRI images of patients and manually built the tailored NPC image dataset. We examined the performance of the representative CNN models including shallow CNN, ResNet50, ResNet101, and EfficientNet-B7. By fine-tuning, shallow CNN, ResNet50, ResNet101, and EfficientNet-B7 achieved the precision of 72.2%, 94.4%, 92.6%, and 88.4%, displaying the superiority of deep hierarchical neural networks. Among the examined models, ResNet50 with pre-trained weights demonstrated the best classification performance over other types of CNN with accuracy, precision, and an F1-score of 0.93, 0.94, and 0.93, respectively. The fine-tuned ResNet50 achieved the highest prediction performance and can be used as a potential tool for aiding the diagnosis of NPC tumors.

Background: The new coronavirus disease 2019 (COVID-19) is spreading rapidly around the world. Artificial Intelligence (AI) assisted identification and detection of diseases is an effective method of medical diagnosis. Objectives: To present recent advances in AI-assisted diagnosis of COVID-19, we introduce major aspects of AI in the process of diagnosing COVID-19. Methods: In this paper, we firstly cover the latest collection and processing methods of datasets of COVID-19. The processing methods mainly include building public datasets, transfer learning, unsupervised learning and weakly supervised learning, semi-supervised learning methods and so on. Secondly, we introduce the algorithm application and evaluation metrics of AI in medical imaging segmentation and automatic screening. Then, we introduce the quantification and severity assessment of infection in COVID-19 patients based on image segmentation and automatic screening. Finally, we analyze and point out the current AI-assisted diagnosis of COVID-19 problems, which may provide useful clues for future work. Conclusion: AI is critical for COVID-19 diagnosis. Combining chest imaging with AI can not only save time and effort, but also provide more accurate and efficient medical diagnosis results.

With the unprecedented developments in deep learning, automatic segmentation of main abdominal organs seems to be a solved problem as state-of-the-art (SOTA) methods have achieved comparable results with inter-rater variability on many benchmark datasets. However, most of the existing abdominal datasets only contain single-center, single-phase, single-vendor, or single-disease cases, and it is unclear whether the excellent performance can generalize on diverse datasets. This paper presents a large and diverse abdominal CT organ segmentation dataset, termed AbdomenCT-1K, with more than 1000 (1K) CT scans from 12 medical centers, including multi-phase, multi-vendor, and multi-disease cases. Furthermore, we conduct a large-scale study for liver, kidney, spleen, and pancreas segmentation and reveal the unsolved segmentation problems of the SOTA methods, such as the limited generalization ability on distinct medical centers, phases, and unseen diseases. To advance the unsolved problems, we further build four organ segmentation benchmarks for fully supervised, semi-supervised, weakly supervised, and continual learning, which are currently challenging and active research topics. Accordingly, we develop a simple and effective method for each benchmark, which can be used as out-of-the-box methods and strong baselines. We believe the AbdomenCT-1K dataset will promote future in-depth research towards clinical applicable abdominal organ segmentation methods.