Motivation: In recent years, the development of natural language process (NLP) technologies and deep learning hardware has led to significant improvement in large language models (LLMs). The ChatGPT, the state-of-the-art LLM built on GPT-3.5 and GPT-4, shows excellent capabilities in general language understanding and reasoning. Researchers also tested the GPTs on a variety of NLP-related tasks and benchmarks and got excellent results. With exciting performance on daily chat, researchers began to explore the capacity of ChatGPT on expertise that requires professional education for human and we are interested in the biomedical domain. Results: To evaluate the performance of ChatGPT on biomedical-related tasks, this article presents a comprehensive benchmark study on the use of ChatGPT for biomedical corpus, including article abstracts, clinical trials description, biomedical questions, and so on. Typical NLP tasks like named entity recognization, relation extraction, sentence similarity, question and answering, and document classification are included. Overall, ChatGPT got a BLURB score of 58.50 while the state-of-the-art model had a score of 84.30. Through a series of experiments, we demonstrated the effectiveness and versatility of ChatGPT in biomedical text understanding, reasoning and generation, and the limitation of ChatGPT build on GPT-3.5. Availability and implementation: All the datasets are available from BLURB benchmark https://microsoft.github.io/BLURB/index.html. The prompts are described in the article.

Spatial structures of proteins are closely related to protein functions. Integrating protein structures improves the performance of protein–protein interaction (PPI) prediction. However, the limited quantity of known protein structures restricts the application of structure-based prediction methods. Utilizing the predicted protein structure information is a promising method to improve the performance of sequence-based prediction methods. We propose a novel end-to-end framework, TAGPPI, to predict PPIs using protein sequence alone. TAGPPI extracts multi-dimensional features by employing 1D convolution operation on protein sequences and graph learning method on contact maps constructed from AlphaFold. A contact map contains abundant spatial structure information, which is difficult to obtain from 1D sequence data directly. We further demonstrate that the spatial information learned from contact maps improves the ability of TAGPPI in PPI prediction tasks. We compare the performance of TAGPPI with those of nine state-of-the-art sequence-based methods, and TAGPPI outperforms such methods in all metrics. To the best of our knowledge, this is the first method to use the predicted protein topology structure graph for sequence-based PPI prediction. More importantly, our proposed architecture could be extended to other prediction tasks related to proteins.

Background: Apolipoprotein E (APOE) ε4 is the single most important genetic risk factor for cognitive impairment and Alzheimer disease (AD), while lifestyle factors such as smoking, drinking, diet, and physical activity also have impact on cognition. The goal of the study is to investigate whether the association between lifestyle and cognition varies by APOE genotype among the oldest old. Methods and findings: We used the cross-sectional data including 6,160 oldest old (aged 80 years old or older) from the genetic substudy of the Chinese Longitudinal Healthy Longevity Survey (CLHLS) which is a national wide cohort study that began in 1998 with follow-up surveys every 2–3 years. Cognitive impairment was defined as a Mini-Mental State Examination (MMSE) score less than 18. Healthy lifestyle profile was classified into 3 groups by a composite measure including smoking, alcohol consumption, dietary pattern, physical activity, and body weight. APOE genotype was categorized as APOE ε4 carriers versus noncarriers. We examined the associations of cognitive impairment with lifestyle profile and APOE genotype using multivariable logistic regressions, controlling for age, sex, education, marital status, residence, disability, and numbers of chronic conditions. The mean age of our study sample was 90.1 (standard deviation [SD], 7.2) years (range 80–113); 57.6% were women, and 17.5% were APOE ε4 carriers. The mean MMSE score was 21.4 (SD: 9.2), and 25.0% had cognitive impairment. Compared with those with an unhealthy lifestyle, participants with intermediate and healthy lifestyle profiles were associated with 28% (95% confidence interval [CI]: 16%–38%, P < 0.001) and 55% (95% CI: 44%–64%, P < 0.001) lower adjusted odds of cognitive impairment. Carrying the APOE ε4 allele was associated with 17% higher odds (95% CI: 1%–31%, P = 0.042) of being cognitively impaired in the adjusted model. The association between lifestyle profiles and cognitive function did not vary significantly by APOE ε4 genotype (noncarriers: 0.47 [0.37–0.60] healthy versus unhealthy; carriers: 0.33 [0.18–0.58], P for interaction = 0.30). The main limitation was the lifestyle measurements were self-reported and were nonspecific. Generalizability of the findings is another limitation because the study sample was from the oldest old in China, with unique characteristics such as low body weight compared to populations in high-income countries. Conclusions: In this study, we observed that healthier lifestyle was associated with better cognitive function among the oldest old regardless of APOE genotype. Our findings may inform the cognitive outlook for those oldest old with high genetic risk of cognitive impairment.

Informative representation of molecules is a crucial prerequisite in AI-driven drug design and discovery. Pharmacophore information including functional groups and chemical reactions can indicate molecular properties, which have not been fully exploited by prior atom-based molecular graph representation. To obtain a more informative representation of molecules for better molecule property prediction, we propose the Pharmacophoric-constrained Heterogeneous Graph Transformer (PharmHGT). We design a pharmacophoric-constrained multi-views molecular representation graph, enabling PharmHGT to extract vital chemical information from functional substructures and chemical reactions. With a carefully designed pharmacophoric-constrained multi-view molecular representation graph, PharmHGT can learn more chemical information from molecular functional substructures and chemical reaction information. Extensive downstream experiments prove that PharmHGT achieves remarkably superior performance over the state-of-the-art models the performance of our model is up to 1.55% in ROC-AUC and 0.272 in RMSE higher than the best baseline model) on molecular properties prediction. The ablation study and case study show that our proposed molecular graph representation method and heterogeneous graph transformer model can better capture the pharmacophoric structure and chemical information features. Further visualization studies also indicated a better representation capacity achieved by our model.

The drug repurposing of known approved drugs (e.g., lopinavir/ritonavir) has failed to treat SARS-CoV-2-infected patients. Therefore, it is important to generate new chemical entities against this virus. As a critical enzyme in the lifecycle of the coronavirus, the 3C-like main protease (3CLpro or Mpro) is the most attractive target for antiviral drug design. Based on a recently solved structure (PDB ID: 6LU7), we developed a novel advanced deep Q-learning network with a fragment-based drug design (ADQN–FBDD) for generating potential lead compounds targeting SARS-CoV-2 3CLpro. We obtained a series of derivatives from the lead compounds based on our structure-based optimization policy (SBOP). All of the 47 lead compounds obtained directly with our AI model and related derivatives based on the SBOP are accessible in our molecular library. These compounds can be used as potential candidates by researchers to develop drugs against SARS-CoV-2.

Prediction of the drug-target affinity (DTA) plays an important role in drug discovery. Existing deep learning methods for DTA prediction typically leverage a single modality, namely simplified molecular input line entry specification (SMILES) or amino acid sequence to learn representations. SMILES or amino acid sequences can be encoded into different modalities. Multimodality data provide different kinds of information , with complementary roles for DTA prediction. We propose Modality-DTA, a novel deep learning method for DTA prediction that leverages the multimodality of drugs and targets. A group of backward propagation neural networks is applied to ensure the completeness of the reconstruction process from the latent feature representation to original multimodality data. The tag between the drug and target is used to reduce the noise information in the latent representation from multimodality data. Experiments on three benchmark datasets show that our Modality-DTA outperforms existing methods in all metrics. Modality-DTA reduces the mean square error by 15.7% and improves the area under the precisionrecall curve by 12.74% in the Davis dataset. We further find that the drug modality Morgan fingerprint and the target modality generated by one-hot-encoding play the most significant roles. To the best of our knowledge, Modality-DTA is the first method to explore multimodality for DTA prediction.