Fine-Tuning Cytokine Signals.

Cytokines are secreted or otherwise released polypeptide factors that exert autocrine and/or paracrine actions, with most cytokines acting in the immune and/or hematopoietic system. They are typically pleiotropic, controlling development, cell growth, survival, and/or differentiation. Correspondingly, cytokines are clinically important, and augmenting or attenuating cytokine signals can have deleterious or therapeutic effects. Besides physiological fine-tuning of cytokine signals, altering the nature or potency of the signal can be important in pathophysiological responses and can also provide novel therapeutic approaches. Here, we give an overview of cytokines, their signaling and actions, and the physiological mechanisms and pharmacologic strategies to fine-tune their actions. In particular, the differential utilization of STAT proteins by a single cytokine or by different cytokines and STAT dimerization versus tetramerization are physiological mechanisms of fine-tuning, whereas anticytokine and anticytokine receptor antibodies and cytokines with altered activities, including cytokine superagonists, partial agonists, and antagonists, represent new ways of fine-tuning cytokine signals.

Democratizing protein language models with parameter-efficient fine-tuning.

Proteomics has been revolutionized by large protein language models (PLMs), which learn unsupervised representations from large corpora of sequences. These models are typically fine-tuned in a supervised setting to adapt the model to specific downstream tasks. However, the computational and memory footprint of fine-tuning (FT) large PLMs presents a barrier for many research groups with limited computational resources. Natural language processing has seen a similar explosion in the size of models, where these challenges have been addressed by methods for parameter-efficient fine-tuning (PEFT). In this work, we introduce this paradigm to proteomics through leveraging the parameter-efficient method LoRA and training new models for two important tasks: predicting protein-protein interactions (PPIs) and predicting the symmetry of homooligomer quaternary structures. We show that these approaches are competitive with traditional FT while requiring reduced memory and substantially fewer parameters. We additionally show that for the PPI prediction task, training only the classification head also remains competitive with full FT, using five orders of magnitude fewer parameters, and that each of these methods outperform state-of-the-art PPI prediction methods with substantially reduced compute. We further perform a comprehensive evaluation of the hyperparameter space, demonstrate that PEFT of PLMs is robust to variations in these hyperparameters, and elucidate where best practices for PEFT in proteomics differ from those in natural language processing. All our model adaptation and evaluation code is available open-source at https://github.com/microsoft/peft_proteomics. Thus, we provide a blueprint to democratize the power of PLM adaptation to groups with limited computational resources.

VISH-Pred: an ensemble of fine-tuned ESM models for protein toxicity prediction.

Peptide- and protein-based therapeutics are becoming a promising treatment regimen for myriad diseases. Toxicity of proteins is the primary hurdle for protein-based therapies. Thus, there is an urgent need for accurate in silico methods for determining toxic proteins to filter the pool of potential candidates. At the same time, it is imperative to precisely identify non-toxic proteins to expand the possibilities for protein-based biologics. To address this challenge, we proposed an ensemble framework, called VISH-Pred, comprising models built by fine-tuning ESM2 transformer models on a large, experimentally validated, curated dataset of protein and peptide toxicities. The primary steps in the VISH-Pred framework are to efficiently estimate protein toxicities taking just the protein sequence as input, employing an under sampling technique to handle the humongous class-imbalance in the data and learning representations from fine-tuned ESM2 protein language models which are then fed to machine learning techniques such as Lightgbm and XGBoost. The VISH-Pred framework is able to correctly identify both peptides/proteins with potential toxicity and non-toxic proteins, achieving a Matthews correlation coefficient of 0.737, 0.716 and 0.322 and F1-score of 0.759, 0.696 and 0.713 on three non-redundant blind tests, respectively, outperforming other methods by over $10\%$ on these quality metrics. Moreover, VISH-Pred achieved the best accuracy and area under receiver operating curve scores on these independent test sets, highlighting the robustness and generalization capability of the framework. By making VISH-Pred available as an easy-to-use web server, we expect it to serve as a valuable asset for future endeavors aimed at discerning the toxicity of peptides and enabling efficient protein-based therapeutics.

Peptide-binding specificity prediction using fine-tuned protein structure prediction networks.

Peptide-binding proteins play key roles in biology, and predicting their binding specificity is a long-standing challenge. While considerable protein structural information is available, the most successful current methods use sequence information alone, in part because it has been a challenge to model the subtle structural changes accompanying sequence substitutions. Protein structure prediction networks such as AlphaFold model sequence-structure relationships very accurately, and we reasoned that if it were possible to specifically train such networks on binding data, more generalizable models could be created. We show that placing a classifier on top of the AlphaFold network and fine-tuning the combined network parameters for both classification and structure prediction accuracy leads to a model with strong generalizable performance on a wide range of Class I and Class II peptide-MHC interactions that approaches the overall performance of the state-of-the-art NetMHCpan sequence-based method. The peptide-MHC optimized model shows excellent performance in distinguishing binding and non-binding peptides to SH3 and PDZ domains. This ability to generalize well beyond the training set far exceeds that of sequence-only models and should be particularly powerful for systems where less experimental data are available.

Alternative splicing of lncRNA LAIR fine-tunes the regulation of neighboring yield-related gene LRK1 expression in rice.

The diversity in alternative splicing of long noncoding RNAs (lncRNAs) poses a challenge for functional annotation of lncRNAs. Moreover, little is known on the effects of alternatively spliced lncRNAs on crop yield. In this study, we cloned nine isoforms resulting from the alternative splicing of the lncRNA LAIR in rice. The LAIR isoforms are generated via alternative 5'/3' splice sites and different combinations of specific introns. All LAIR isoforms activate the expression of the neighboring LRK1 gene and enhance yield-related rice traits. In addition, there are slight differences in the binding ability of LAIR isoforms to the epigenetic modification-related proteins OsMOF and OsWDR5, which affect the enrichment of H4K16ac and H3K4me3 at the LRK1 locus, and consequently fine-tune the regulation of LRK1 expression and yield-related traits. These differences in binding may be caused by polymorphic changes to the RNA secondary structure resulting from alternative splicing. It was also observed that the composition of LAIR isoforms was sensitive to abiotic stress. These findings suggest that the alternative splicing of LAIR leads to the formation of a functional transcript population that precisely regulates yield-related gene expression, which may be relevant for phenotypic polymorphism-based crop breeding under changing environmental conditions.

Assessing parameter efficient methods for pre-trained language model in annotating scRNA-seq data.

Annotating cell types of single-cell RNA sequencing (scRNA-seq) data is crucial for studying cellular heterogeneity in the tumor microenvironment. Recently, large-scale pre-trained language models (PLMs) have achieved significant progress in cell-type annotation of scRNA-seq data. This approach effectively addresses previous methods' shortcomings in performance and generalization. However, fine-tuning PLMs for different downstream tasks demands considerable computational resources, rendering it impractical. Hence, a new research branch introduces parameter-efficient fine-tuning (PEFT). This involves optimizing a few parameters while leaving the majority unchanged, leading to substantial reductions in computational expenses. Here, we utilize scBERT, a large-scale pre-trained model, to explore the capabilities of three PEFT methods in scRNA-seq cell type annotation. Extensive benchmark studies across several datasets demonstrate the superior applicability of PEFT methods. Furthermore, downstream analysis using models obtained through PEFT showcases their utility in novel cell type discovery and model interpretability for potential marker genes. Our findings underscore the considerable potential of PEFT in PLM-based cell type annotation, presenting novel perspectives for the analysis of scRNA-seq data.

Learning self-supervised molecular representations for drug-drug interaction prediction.

Drug-drug interactions (DDI) are a critical concern in healthcare due to their potential to cause adverse effects and compromise patient safety. Supervised machine learning models for DDI prediction need to be optimized to learn abstract, transferable features, and generalize to larger chemical spaces, primarily due to the scarcity of high-quality labeled DDI data. Inspired by recent advances in computer vision, we present SMR-DDI, a self-supervised framework that leverages contrastive learning to embed drugs into a scaffold-based feature space. Molecular scaffolds represent the core structural motifs that drive pharmacological activities, making them valuable for learning informative representations. Specifically, we pre-trained SMR-DDI on a large-scale unlabeled molecular dataset. We generated augmented views for each molecule via SMILES enumeration and optimized the embedding process through contrastive loss minimization between views. This enables the model to capture relevant and robust molecular features while reducing noise. We then transfer the learned representations for the downstream prediction of DDI. Experiments show that the new feature space has comparable expressivity to state-of-the-art molecular representations and achieved competitive DDI prediction results while training on less data. Additional investigations also revealed that pre-training on more extensive and diverse unlabeled molecular datasets improved the model's capability to embed molecules more effectively. Our results highlight contrastive learning as a promising approach for DDI prediction that can identify potentially hazardous drug combinations using only structural information.

Prediction of potential miRNA-disease associations based on stacked autoencoder.

In recent years, increasing biological experiments and scientific studies have demonstrated that microRNA (miRNA) plays an important role in the development of human complex diseases. Therefore, discovering miRNA-disease associations can contribute to accurate diagnosis and effective treatment of diseases. Identifying miRNA-disease associations through computational methods based on biological data has been proven to be low-cost and high-efficiency. In this study, we proposed a computational model named Stacked Autoencoder for potential MiRNA-Disease Association prediction (SAEMDA). In SAEMDA, all the miRNA-disease samples were used to pretrain a Stacked Autoencoder (SAE) in an unsupervised manner. Then, the positive samples and the same number of selected negative samples were utilized to fine-tune SAE in a supervised manner after adding an output layer with softmax classifier to the SAE. SAEMDA can make full use of the feature information of all unlabeled miRNA-disease pairs. Therefore, SAEMDA is suitable for our dataset containing small labeled samples and large unlabeled samples. As a result, SAEMDA achieved AUCs of 0.9210 and 0.8343 in global and local leave-one-out cross validation. Besides, SAEMDA obtained an average AUC and standard deviation of 0.9102 ± /-0.0029 in 100 times of 5-fold cross validation. These results were better than those of previous models. Moreover, we carried out three case studies to further demonstrate the predictive accuracy of SAEMDA. As a result, 82% (breast neoplasms), 100% (lung neoplasms) and 90% (esophageal neoplasms) of the top 50 predicted miRNAs were verified by databases. Thus, SAEMDA could be a useful and reliable model to predict potential miRNA-disease associations.

An analysis of protein language model embeddings for fold prediction.

The identification of the protein fold class is a challenging problem in structural biology. Recent computational methods for fold prediction leverage deep learning techniques to extract protein fold-representative embeddings mainly using evolutionary information in the form of multiple sequence alignment (MSA) as input source. In contrast, protein language models (LM) have reshaped the field thanks to their ability to learn efficient protein representations (protein-LM embeddings) from purely sequential information in a self-supervised manner. In this paper, we analyze a framework for protein fold prediction using pre-trained protein-LM embeddings as input to several fine-tuning neural network models, which are supervisedly trained with fold labels. In particular, we compare the performance of six protein-LM embeddings: the long short-term memory-based UniRep and SeqVec, and the transformer-based ESM-1b, ESM-MSA, ProtBERT and ProtT5; as well as three neural networks: Multi-Layer Perceptron, ResCNN-BGRU (RBG) and Light-Attention (LAT). We separately evaluated the pairwise fold recognition (PFR) and direct fold classification (DFC) tasks on well-known benchmark datasets. The results indicate that the combination of transformer-based embeddings, particularly those obtained at amino acid level, with the RBG and LAT fine-tuning models performs remarkably well in both tasks. To further increase prediction accuracy, we propose several ensemble strategies for PFR and DFC, which provide a significant performance boost over the current state-of-the-art results. All this suggests that moving from traditional protein representations to protein-LM embeddings is a very promising approach to protein fold-related tasks.

Enhancing recognition and interpretation of functional phenotypic sequences through fine-tuning pre-trained genomic models.

BACKGROUND: Decoding human genomic sequences requires comprehensive analysis of DNA sequence functionality. Through computational and experimental approaches, researchers have studied the genotype-phenotype relationship and generate important datasets that help unravel complicated genetic blueprints. Thus, the recently developed artificial intelligence methods can be used to interpret the functions of those DNA sequences. METHODS: This study explores the use of deep learning, particularly pre-trained genomic models like DNA_bert_6 and human_gpt2-v1, in interpreting and representing human genome sequences. Initially, we meticulously constructed multiple datasets linking genotypes and phenotypes to fine-tune those models for precise DNA sequence classification. Additionally, we evaluate the influence of sequence length on classification results and analyze the impact of feature extraction in the hidden layers of our model using the HERV dataset. To enhance our understanding of phenotype-specific patterns recognized by the model, we perform enrichment, pathogenicity and conservation analyzes of specific motifs in the human endogenous retrovirus (HERV) sequence with high average local representation weight (ALRW) scores. RESULTS: We have constructed multiple genotype-phenotype datasets displaying commendable classification performance in comparison with random genomic sequences, particularly in the HERV dataset, which achieved binary and multi-classification accuracies and F1 values exceeding 0.935 and 0.888, respectively. Notably, the fine-tuning of the HERV dataset not only improved our ability to identify and distinguish diverse information types within DNA sequences but also successfully identified specific motifs associated with neurological disorders and cancers in regions with high ALRW scores. Subsequent analysis of these motifs shed light on the adaptive responses of species to environmental pressures and their co-evolution with pathogens. CONCLUSIONS: These findings highlight the potential of pre-trained genomic models in learning DNA sequence representations, particularly when utilizing the HERV dataset, and provide valuable insights for future research endeavors. This study represents an innovative strategy that combines pre-trained genomic model representations with classical methods for analyzing the functionality of genome sequences, thereby promoting cross-fertilization between genomics and artificial intelligence.

Optimizing large language models in digestive disease: strategies and challenges to improve clinical outcomes.

Large Language Models (LLMs) are transformer-based neural networks with billions of parameters trained on very large text corpora from diverse sources. LLMs have the potential to improve healthcare due to their capability to parse complex concepts and generate context-based responses. The interest in LLMs has not spared digestive disease academics, who have mainly investigated foundational LLM accuracy, which ranges from 25% to 90% and is influenced by the lack of standardized rules to report methodologies and results for LLM-oriented research. In addition, a critical issue is the absence of a universally accepted definition of accuracy, varying from binary to scalar interpretations, often tied to grader expertise without reference to clinical guidelines. We address strategies and challenges to increase accuracy. In particular, LLMs can be infused with domain knowledge using Retrieval Augmented Generation (RAG) or Supervised Fine-Tuning (SFT) with reinforcement learning from human feedback (RLHF). RAG faces challenges with in-context window limits and accurate information retrieval from the provided context. SFT, a deeper adaptation method, is computationally demanding and requires specialized knowledge. LLMs may increase patient quality of care across the field of digestive diseases, where physicians are often engaged in screening, treatment and surveillance for a broad range of pathologies for which in-context learning or SFT with RLHF could improve clinical decision-making and patient outcomes. However, despite their potential, the safe deployment of LLMs in healthcare still needs to overcome hurdles in accuracy, suggesting a need for strategies that integrate human feedback with advanced model training.

Transformers for Molecular Property Prediction: Lessons Learned from the Past Five Years.

Molecular Property Prediction (MPP) is vital for drug discovery, crop protection, and environmental science. Over the last decades, diverse computational techniques have been developed, from using simple physical and chemical properties and molecular fingerprints in statistical models and classical machine learning to advanced deep learning approaches. In this review, we aim to distill insights from current research on employing transformer models for MPP. We analyze the currently available models and explore key questions that arise when training and fine-tuning a transformer model for MPP. These questions encompass the choice and scale of the pretraining data, optimal architecture selections, and promising pretraining objectives. Our analysis highlights areas not yet covered in current research, inviting further exploration to enhance the field's understanding. Additionally, we address the challenges in comparing different models, emphasizing the need for standardized data splitting and robust statistical analysis.

Identification of the causal mutation in early heading mutant of bread wheat (Triticum aestivum L.) using MutMap approach.

In bread wheat (Triticum aestivum L.), fine-tuning the heading time is essential to maximize grain yield. Photoperiod-1 (Ppd-1) and VERNALIZATION 1 (Vrn-1) are major genes affecting photoperiod sensitivity and vernalization requirements, respectively. These genes have predominantly governed heading timing. However, Ppd-1 and Vrn-1 significantly impact heading dates, necessitating another gene that can slightly modify heading dates for fine-tuning. In this study, we developed an early heading mutant from the ethyl methanesulfonate-mutagenized population of the Japanese winter wheat cultivar "Kitahonami." MutMap analysis identified a nonsense mutation in the clock component gene Wheat PHYTOCLOCK 1/LUX ARRHYTHMO (WPCL-D1) as the probable SNP responsible for the early heading mutant on chromosome 3D. Segregation analysis using F2 and F3 populations confirmed that plants carrying the wpcl-D1 allele headed significantly earlier than those with the functional WPCL-D1. The early heading mutant exhibited increased expression levels of Ppd-1 and circadian clock genes, such as WPCL1 and LATE ELONGATED HYPOCOTYL (LHY). Notably, the transcript accumulation levels of Ppd-A1 and Ppd-D1 were influenced by the copy number of the functional WPCL1 gene. These results suggest that a loss-of-function mutation in WPCL-D1 is the causal mutation for the early heading phenotype. Adjusting the functional copy number of WPCL1 will be beneficial in fine-tuning of heading dates. Supplementary Information: The online version contains supplementary material available at 10.1007/s11032-024-01478-5.

Revolutionizing breast ultrasound diagnostics with EfficientNet-B7 and Explainable AI.

Breast cancer is a leading cause of mortality among women globally, necessitating precise classification of breast ultrasound images for early diagnosis and treatment. Traditional methods using CNN architectures such as VGG, ResNet, and DenseNet, though somewhat effective, often struggle with class imbalances and subtle texture variations, leading to reduced accuracy for minority classes such as malignant tumors. To address these issues, we propose a methodology that leverages EfficientNet-B7, a scalable CNN architecture, combined with advanced data augmentation techniques to enhance minority class representation and improve model robustness. Our approach involves fine-tuning EfficientNet-B7 on the BUSI dataset, implementing RandomHorizontalFlip, RandomRotation, and ColorJitter to balance the dataset and improve model robustness. The training process includes early stopping to prevent overfitting and optimize performance metrics. Additionally, we integrate Explainable AI (XAI) techniques, such as Grad-CAM, to enhance the interpretability and transparency of the model's predictions, providing visual and quantitative insights into the features and regions of ultrasound images influencing classification outcomes. Our model achieves a classification accuracy of 99.14%, significantly outperforming existing CNN-based approaches in breast ultrasound image classification. The incorporation of XAI techniques enhances our understanding of the model's decision-making process, thereby increasing its reliability and facilitating clinical adoption. This comprehensive framework offers a robust and interpretable tool for the early detection and diagnosis of breast cancer, advancing the capabilities of automated diagnostic systems and supporting clinical decision-making processes.

Ssu72 phosphatase directly binds to ZAP-70, thereby providing fine-tuning of TCR signaling and preventing spontaneous inflammation.

ZAP-70 is required for the initiation of T cell receptor (TCR) signaling, and Ssu72 is a phosphatase that regulates RNA polymerase II activity in the nucleus. However, the mechanism by which ZAP-70 regulates the fine-tuning of TCR signaling remains elusive. Here, we found that Ssu72 contributed to the fine-tuning of TCR signaling by acting as tyrosine phosphatase for ZAP-70. Affinity purification-mass spectrometry and an in vitro assay demonstrated specific interaction between Ssu72 and ZAP-70 in T cells. Upon TCR stimulation, Ssu72-deficient T cells increased the phosphorylation of ZAP-70 and downstream molecules and exhibited hyperresponsiveness, which was restored by reducing ZAP-70 phosphorylation. In vitro assay demonstrated that recombinant Ssu72 reduced tyrosine phosphorylation of ZAP-70 via phosphatase activity. Cd4-CreSsu72fl/fl mice showed a defect in the thymic development of invariant natural killer T cells and reductions in CD4+ and CD8+ T cell numbers in the periphery but more CD44hiCD62Llo memory T cells and fewer CD44loCD62Lhi naïve T cells, compared with wild-type mice. Furthermore, Cd4-CreSsu72fl/fl mice developed spontaneous inflammation at 6 mo. In conclusion, Ssu72 phosphatase regulates the fine-tuning of TCR signaling by binding to ZAP-70 and regulating its tyrosine phosphorylation, thereby preventing spontaneous inflammation.

Short- and long-term effects of remote synchronous fine-tuning for first-time hearing aid users: a randomised controlled trial.

OBJECTIVE: To evaluate the short- and long-term effect of remote synchronous fine-tuning and follow-up visits on hearing-related problems and hearing aid (HA) benefits for first-time HA users. DESIGN: A randomised controlled trial. STUDY SAMPLE: Patients at public tax-funded HA clinics in Sweden due for aural rehabilitation (AR) were randomised to either an intervention group (n = 33) or a control group (n = 36). Both groups completed the conventional AR process, but the intervention group received synchronous remote fine-tuning of HAs and online follow-up visits. Outcome measures were used before and after intervention, and 6 months and 1 year after intervention. RESULTS: Both groups improved hearing-related problems measured with the Hearing Handicap Inventory for the Elderly/Adults over time, and no significant differences were found between the groups. Such improvements were also found for the Abbreviated Profile of Hearing Aid Benefit except for the subscale aversiveness. Both groups decreased the use of HAs in hours/day over time. The intervention group reported significant improvements in activity limitation when measured directly after the intervention, compared to the control group. CONCLUSION: Synchronous remote fine-tuning and follow-ups for first-time HA users did not influence the outcomes of hearing-related problems and HA benefits differently from standard care at our clinic.

Accurate and robust auto-segmentation of head and neck organ-at-risks based on a novel CNN fine-tuning workflow.

PURPOSE: Obvious inconsistencies in auto-segmentations exist among various AI software. In this study, we have developed a novel convolutional neural network (CNN) fine-tuning workflow to achieve precise and robust localized segmentation. METHODS: The datasets include Hubei Cancer Hospital dataset, Cetuximab Head and Neck Public Dataset, and Québec Public Dataset. Seven organs-at-risks (OARs), including brain stem, left parotid gland, esophagus, left optic nerve, optic chiasm, mandible, and pharyngeal constrictor, were selected. The auto-segmentation results from four commercial AI software were first compared with the manual delineations. Then a new multi-scale lightweight residual CNN model with an attention module (named as HN-Net) was trained and tested on 40 samples and 10 samples from Hubei Cancer Hospital, respectively. To enhance the network's accuracy and generalization ability, the fine-tuning workflow utilized an uncertainty estimation method for automatic selection of candidate samples of worthiness from Cetuximab Head and Neck Public Dataset for further training. The segmentation performances were evaluated on the Hubei Cancer Hospital dataset and/or the entire Québec Public Dataset. RESULTS: A maximum difference of 0.13 and 0.7 mm in average Dice value and Hausdorff distance value for the seven OARs were observed by four AI software. The proposed HN-Net achieved an average Dice value of 0.14 higher than that of the AI software, and it also outperformed other popular CNN models (HN-Net: 0.79, U-Net: 0.78, U-Net++: 0.78, U-Net-Multi-scale: 0.77, AI software: 0.65). Additionally, the HN-Net fine-tuning workflow by using the local datasets and external public datasets further improved the automatic segmentation with the average Dice value by 0.02. CONCLUSION: The delineations of commercial AI software need to be carefully reviewed, and localized further training is necessary for clinical practice. The proposed fine-tuning workflow could be feasibly adopted to implement an accurate and robust auto-segmentation model by using local datasets and external public datasets.

Multi-Task Scenario Encrypted Traffic Classification and Parameter Analysis.

The widespread use of encrypted traffic poses challenges to network management and network security. Traditional machine learning-based methods for encrypted traffic classification no longer meet the demands of management and security. The application of deep learning technology in encrypted traffic classification significantly improves the accuracy of models. This study focuses primarily on encrypted traffic classification in the fields of network analysis and network security. To address the shortcomings of existing deep learning-based encrypted traffic classification methods in terms of computational memory consumption and interpretability, we introduce a Parameter-Efficient Fine-Tuning method for efficiently tuning the parameters of an encrypted traffic classification model. Experimentation is conducted on various classification scenarios, including Tor traffic service classification and malicious traffic classification, using multiple public datasets. Fair comparisons are made with state-of-the-art deep learning model architectures. The results indicate that the proposed method significantly reduces the scale of fine-tuning parameters and computational resource usage while achieving performance comparable to that of the existing best models. Furthermore, we interpret the learning mechanism of encrypted traffic representation in the pre-training model by analyzing the parameters and structure of the model. This comparison validates the hypothesis that the model exhibits hierarchical structure, clear organization, and distinct features.

DriveLLaVA: Human-Level Behavior Decisions via Vision Language Model.

Human-level driving is the ultimate goal of autonomous driving. As the top-level decision-making aspect of autonomous driving, behavior decision establishes short-term driving behavior strategies by evaluating road structures, adhering to traffic rules, and analyzing the intentions of other traffic participants. Existing behavior decisions are primarily implemented based on rule-based methods, exhibiting insufficient generalization capabilities when faced with new and unseen driving scenarios. In this paper, we propose a novel behavior decision method that leverages the inherent generalization and commonsense reasoning abilities of visual language models (VLMs) to learn and simulate the behavior decision process in human driving. We constructed a novel instruction-following dataset containing a large number of image-text instructions paired with corresponding driving behavior labels, to support the learning of the Drive Large Language and Vision Assistant (DriveLLaVA) and enhance the transparency and interpretability of the entire decision process. DriveLLaVA is fine-tuned on this dataset using the Low-Rank Adaptation (LoRA) approach, which efficiently optimizes the model parameter count and significantly reduces training costs. We conducted extensive experiments on a large-scale instruction-following dataset, and compared with state-of-the-art methods, DriveLLaVA demonstrated excellent behavior decision performance. DriveLLaVA is capable of handling various complex driving scenarios, showing strong robustness and generalization abilities.

Adapt-cMolGPT: A Conditional Generative Pre-Trained Transformer with Adapter-Based Fine-Tuning for Target-Specific Molecular Generation.

Small-molecule drug design aims to generate compounds that target specific proteins, playing a crucial role in the early stages of drug discovery. Recently, research has emerged that utilizes the GPT model, which has achieved significant success in various fields to generate molecular compounds. However, due to the persistent challenge of small datasets in the pharmaceutical field, there has been some degradation in the performance of generating target-specific compounds. To address this issue, we propose an enhanced target-specific drug generation model, Adapt-cMolGPT, which modifies molecular representation and optimizes the fine-tuning process. In particular, we introduce a new fine-tuning method that incorporates an adapter module into a pre-trained base model and alternates weight updates by sections. We evaluated the proposed model through multiple experiments and demonstrated performance improvements compared to previous models. In the experimental results, Adapt-cMolGPT generated a greater number of novel and valid compounds compared to other models, with these generated compounds exhibiting properties similar to those of real molecular data. These results indicate that our proposed method is highly effective in designing drugs targeting specific proteins.