SAM-DTA: a sequence-agnostic model for drug-target binding affinity prediction.

Drug-target binding affinity prediction is a fundamental task for drug discovery and has been studied for decades. Most methods follow the canonical paradigm that processes the inputs of the protein (target) and the ligand (drug) separately and then combines them together. In this study we demonstrate, surprisingly, that a model is able to achieve even superior performance without access to any protein-sequence-related information. Instead, a protein is characterized completely by the ligands that it interacts. Specifically, we treat different proteins separately, which are jointly trained in a multi-head manner, so as to learn a robust and universal representation of ligands that is generalizable across proteins. Empirical evidences show that the novel paradigm outperforms its competitive sequence-based counterpart, with the Mean Squared Error (MSE) of 0.4261 versus 0.7612 and the R-Square of 0.7984 versus 0.6570 compared with DeepAffinity. We also investigate the transfer learning scenario where unseen proteins are encountered after the initial training, and the cross-dataset evaluation for prospective studies. The results reveals the robustness of the proposed model in generalizing to unseen proteins as well as in predicting future data. Source codes and data are available at https://github.com/huzqatpku/SAM-DTA.

Simple, Efficient, and Scalable Structure-Aware Adapter Boosts Protein Language Models.

Fine-tuning pretrained protein language models (PLMs) has emerged as a prominent strategy for enhancing downstream prediction tasks, often outperforming traditional supervised learning approaches. As a widely applied powerful technique in natural language processing, employing parameter-efficient fine-tuning techniques could potentially enhance the performance of PLMs. However, the direct transfer to life science tasks is nontrivial due to the different training strategies and data forms. To address this gap, we introduce SES-Adapter, a simple, efficient, and scalable adapter method for enhancing the representation learning of PLMs. SES-Adapter incorporates PLM embeddings with structural sequence embeddings to create structure-aware representations. We show that the proposed method is compatible with different PLM architectures and across diverse tasks. Extensive evaluations are conducted on 2 types of folding structures with notable quality differences, 9 state-of-the-art baselines, and 9 benchmark data sets across distinct downstream tasks. Results show that compared to vanilla PLMs, SES-Adapter improves downstream task performance by a maximum of 11% and an average of 3%, with significantly accelerated convergence speed by a maximum of 1034% and an average of 362%, the training efficiency is also improved by approximately 2 times. Moreover, positive optimization is observed even with low-quality predicted structures. The source code for SES-Adapter is available at https://github.com/tyang816/SES-Adapter.

Embracing Multiheterogeneity and Privacy Security Simultaneously: A Dynamic Privacy-Aware Federated Reinforcement Learning Approach.

With growing demand for privacy-preserving reinforcement learning (RL) applications, federated RL (FRL) has emerged as a potential solution. However, existing FRL methods struggle with multiple sources of heterogeneity, while lacking robust privacy guarantees. In this study, we propose DPA-FedRL, the dynamic privacy-aware FRL framework, to simultaneously mitigate both issues. First, we innovatively put forward the concept of "multiheterogeneity" and embed the environmental heterogeneity into agents' state representations. Next, to ensure privacy during model aggregation, we incorporate a differentially private mechanism in form of Gaussian noise and modify its global sensitivity, tailored to suit FRL's unique characteristics. Encouragingly, our approach dynamically allocates privacy budget based on heterogeneity levels, which strikes a balance between privacy and utility. From the theoretical perspective, we give rigorous convergence, privacy, and sensitivity guarantees for our proposed method. Through extensive experiments on diverse datasets, we demonstrate that DPA-FedRL surpasses state-of-the-art approaches (PPO-DP-SGD, PAvg, and QAvg) in some highly heterogeneous environments. Notably, our novel privacy attack simulations enable quantitative privacy assessment, validating that DPA-FedRL offers over 1.359 × stronger protection than baselines.

Non-negligible inhibition effect of microcystin-LR biodegradation products target to protein phosphatase 2A.

Though biodegradation is an important regulation pathway for microcystins (MCs) pollution, more consideration needs to be given to the potential risk associated with related biodegradation products (MC-BDPs). In this work, typical MCLR-BDPs were prepared and their toxicity was evaluated by protein phosphatases (PPs) inhibition assay. Results showed the initial ring opening of MCLR played a crucial role in detoxification. However, partial MCLR-BDPs still retained the critical structures and thus exhibited certain toxicity (2.8-43.5% of MCLR). With the aid of molecular simulation, the mechanism for the potential toxicity of BDPs targeting PP2A was elucidated. The initial ring opening made the loss of hydrogen bond Leu2←Arg89, and pi-H bond Adda5-His191, which was responsible for the significant reduction in the toxicity of MCLR-BDP. However, the key hydrogen bonds MeAsp3←Arg89, Glu6←Arg89, Adda5←Asn117, Adda5←His118, Arg4→Pro213, Arg4←Arg214, Ala1←Arg268, and Mdha7←Arg268, metal bond Glu6-Mn12+, and ionic bonds Glu6-Arg89, and Glu6-Mn22+ were preserved in varying degrees. Above preserved interactions maintained the interactions between PP2A and Mn2+ ions (reducing the exposure of Mn2+ ions). Above preserved interactions also hindered the combination of phosphate groups to Arg214 residual and thus exhibited potential toxicity.

PETA: evaluating the impact of protein transfer learning with sub-word tokenization on downstream applications.

Protein language models (PLMs) play a dominant role in protein representation learning. Most existing PLMs regard proteins as sequences of 20 natural amino acids. The problem with this representation method is that it simply divides the protein sequence into sequences of individual amino acids, ignoring the fact that certain residues often occur together. Therefore, it is inappropriate to view amino acids as isolated tokens. Instead, the PLMs should recognize the frequently occurring combinations of amino acids as a single token. In this study, we use the byte-pair-encoding algorithm and unigram to construct advanced residue vocabularies for protein sequence tokenization, and we have shown that PLMs pre-trained using these advanced vocabularies exhibit superior performance on downstream tasks when compared to those trained with simple vocabularies. Furthermore, we introduce PETA, a comprehensive benchmark for systematically evaluating PLMs. We find that vocabularies comprising 50 and 200 elements achieve optimal performance. Our code, model weights, and datasets are available at https://github.com/ginnm/ProteinPretraining . SCIENTIFIC CONTRIBUTION: This study introduces advanced protein sequence tokenization analysis, leveraging the byte-pair-encoding algorithm and unigram. By recognizing frequently occurring combinations of amino acids as single tokens, our proposed method enhances the performance of PLMs on downstream tasks. Additionally, we present PETA, a new comprehensive benchmark for the systematic evaluation of PLMs, demonstrating that vocabularies of 50 and 200 elements offer optimal performance.

MedChatZH: A tuning LLM for traditional Chinese medicine consultations.

Generative Large Language Models (LLMs) have achieved significant success in various natural language processing tasks, including Question-Answering (QA) and dialogue systems. However, most models are trained on English data and lack strong generalization in providing answers in Chinese. This limitation is especially evident in specialized domains like traditional Chinese medical QA, where performance suffers due to the absence of fine-tuning and high-quality datasets. To address this, we introduce MedChatZH, a dialogue model optimized for Chinese medical QA based on transformer decoder with LLaMA architecture. Continued pre-training on a curated corpus of Chinese medical books is followed by fine-tuning with a carefully selected medical instruction dataset, resulting in MedChatZH outperforming several Chinese dialogue baselines on a real-world medical dialogue dataset. Our model, code, and dataset are publicly available on GitHub (https://github.com/tyang816/MedChatZH) to encourage further research in traditional Chinese medicine and LLMs.

Regulation Effectiveness and Mechanism of Biotransformation Pathway on the Toxicity of Microcystin-LR Target to Protein Phosphatase 2A.

Biotransformation is recognized as a potential pathway to regulate the environmental risk of microcystins (MCs). To explore the regulation effectiveness and mechanism of the biotransformation pathway, six typical MCLR-biotransformation products (MCLR-BTPs) were prepared, and their inhibition effects on protein phosphatase 2A (PP2A) were evaluated. The inhibition effects of the MCLR-BTPs generally decreased with the increase in biothiol molecular weights and polarity, indicating that biotransformation was an effective pathway through which to regulate MCLR toxicity. To further explore the regulation mechanism, the key interaction processes between the MCLR/MCLR-BTPs and the PP2A were explored by homology modeling and molecular docking. The introduced biothiols blocked the covalent binding of Mdha7 to Cys269 but strengthened the hydrogen bond "Mdha7"→Arg268. The changed "Mdha7" intervened the combination of MCLR-BTPs to PP2A by weakening the hydrogen bonds Arg4←Arg214, Arg4→Pro213, Adda5←His118, and Ala1←Arg268, and the ionic bond Glu6-Mn12+. The weakening combination of the MCLR-BTPs to PP2A further attenuated the interactions between the conserved domain and the Mn2+ ions (including the ionic bonds Asp57-Mn12+ and Asp85-Mn12+ and the metal bonds Asp57-Mn12+ and Asn117-Mn12+) and increased the exposure of the Mn2+ ions. Meanwhile, the weakened hydrogen bond Arg4←Arg214 facilitated the combination of the phosphate group to Arg214 (with increased exposure). In this way, the catalytic activity of the PP2A was restored.

Insight into the Molecular Mechanism for the Discrepant Inhibition of Microcystins (MCLR, LA, LF, LW, LY) on Protein Phosphatase 2A.

Microcystins (MCs) exhibit diversified inhibition effects on protein phosphatases (PPs) due to their structural differences. To fully evaluate the potential mechanism for the discrepant inhibition effects, the five most frequent MCs with varying residues at position Z4 were selected as the tested toxins. Their inhibition sequence on PP2A was detected as follows: MCLR > MCLW > MCLA > MCLF > MCLY. Combined with homology modeling and molecular docking technology, the major interaction parameters between the MCs and PP2A were obtained. The correlation analysis for the major interaction parameters and inhibition effects showed that the hydrophobicity of Z4 had an important influence on the interaction of the MCs to PP2A. The introduction of hydrophobic Z4 directly weakened hydrogen bonds Z4→Pro213 and Z4←Arg214, indirectly weakened hydrogen bonds Adda5←Asn117, Glu6←Arg89, and MeAsp3←Arg89, but indirectly enhanced ionic bonds Glu6←Arg89, Glu6-Mn12+, and Glu6-Mn22+. In this way, the combination of the MCs with PP2A was blocked, and thus, the interactions between PP2A and the Mn2+ ions (in the catalytic center) were further affected; metal bonds Asp85-Mn12+ and Asp85-Mn22+ were weakened, while metal bond His241-Mn12+ was enhanced. As a result, the interactions in the catalytic center were inhibited to varying degrees, resulting in the reduced toxicity of MCs.

Mechanism for the Potential Inhibition Effect of Microcystin-LR Disinfectant By-Products on Protein Phosphatase 2A.

The secondary contamination of microcystin disinfection by-products (MC-DBPs) is of concern due to the residual structure similar to their original toxin. Based on identification and preparation, the potential inhibition effect of typical MCLR-DBPs (associated with the oxidation of Adda5) on PP2A was confirmed in the sequence of MCLR > P1 > P4 > P3 ≈ P2 > P7 ≈ P6 ≈ P5 > P8. To elucidate the molecular mechanism underlying the inhibition effect, the interaction models for typical MCLR-DBPs and PP2A were constructed using a modeling-based-on-ligand-similarity approach, and the candidate interaction parameters between typical MCLR-DBPs and PP2A were obtained by molecular docking. By analyzing the correlation between inhibition data and candidate interaction parameters, the key interaction parameters were filtered as hydrogen bonds "Adda5"←Asn117, "Adda5"←His118, MeAsp3←Arg89, Arg4←Arg214, Arg4→Pro213; ionic bonds Glu6-Arg89, Asp85-Mn12+, Asp57-Mn22+; and metal bonds Glu6-Mn12+, Glu6-Mn22+. With the gradual intensification of chlorination, Adda5 was destroyed to varying degrees. The key interactions changed correspondingly, resulting in the discrepant inhibition effects of typical MCLR-DBPs on PP2A.

A pipeline to evaluate the discrepant interactions between typical nitrogenous disinfection byproduct haloacetonitriles and human hemoglobin.

To evaluate the interaction between haloacetonitriles (HANs) and human hemoglobin (Hb), a pipeline was established based on fluorescence spectra, mass spectra and molecular docking. Fluorescence spectra analysis showed the fluorescence of Hb was statically quenched by HANs in the sequence of TCAN > DBAN > DCAN > IAN > BAN > CAN. HANs could combine to multiple surface sites of Hb accounting for "hydrogen bonds" and "van der Waals forces". The high-resolution mass spectra analysis for Hb with and without HANs further confirmed the formation of multiple HAN-Hb complexes with different conversion rates. With the assistance of MOE molecule docking, the potential combination sites and related interactions parameters between HANs and Hb were filtrated. By analyzing the correlations between the candidate interactions parameters and fluorescence quenching constants/MS conversion rates, the combination sites of HANs were fixed at Asp126 (α1/α2), Lys127 (α1/α2) in the form of "hydrogen bonds" X â†’ Asp126 (α1/α2), N â†’ Lys127 (α1/α2). In this way, the potential interactions between HANs and Hb were effectively evaluated.

Preparation and properties of novel hydrogels from oxidized konjac glucomannan cross-linked chitosan for in vitro drug delivery.

In this paper, a novel composite hydrogel was prepared by the use of dialdehyde konjac glucomannan (DAK) as macromolecular cross-linking agent for chitosan (CS). This biocompatible material cross-links and gels in minutes. The structure and morphology were characterized by various analyses. The results indicate that the hydrogels formed through the Schiff-base reaction between the amino groups of CS chains and the aldehyde groups of DAK. The cross-link density (rho(x)) increases with the enhancement of DAK content in hydrogels, while equilibrium swelling ratio (SR) and the average molecular weight between cross-links (Mc) value decrease. Drug release was evaluated by varying the pH of the release medium, reversed dependence of release rate on the equilibrium SR of hydrogel indicated that drug release may be impeded by the association of drug with the polymer. Importantly, this process offers an entirely new window of materials preparation when compared with the traditional preparation of CS-based hydrogels with small molecules cross-linking agent.