Learning the shape of protein microenvironments with a holographic convolutional neural network.

Proteins play a central role in biology from immune recognition to brain activity. While major advances in machine learning have improved our ability to predict protein structure from sequence, determining protein function from its sequence or structure remains a major challenge. Here, we introduce holographic convolutional neural network (H-CNN) for proteins, which is a physically motivated machine learning approach to model amino acid preferences in protein structures. H-CNN reflects physical interactions in a protein structure and recapitulates the functional information stored in evolutionary data. H-CNN accurately predicts the impact of mutations on protein stability and binding of protein complexes. Our interpretable computational model for protein structure-function maps could guide design of novel proteins with desired function.

ADH-Enhancer: an attention-based deep hybrid framework for enhancer identification and strength prediction.

Enhancers play an important role in the process of gene expression regulation. In DNA sequence abundance or absence of enhancers and irregularities in the strength of enhancers affects gene expression process that leads to the initiation and propagation of diverse types of genetic diseases such as hemophilia, bladder cancer, diabetes and congenital disorders. Enhancer identification and strength prediction through experimental approaches is expensive, time-consuming and error-prone. To accelerate and expedite the research related to enhancers identification and strength prediction, around 19 computational frameworks have been proposed. These frameworks used machine and deep learning methods that take raw DNA sequences and predict enhancer's presence and strength. However, these frameworks still lack in performance and are not useful in real time analysis. This paper presents a novel deep learning framework that uses language modeling strategies for transforming DNA sequences into statistical feature space. It applies transfer learning by training a language model in an unsupervised fashion by predicting a group of nucleotides also known as k-mers based on the context of existing k-mers in a sequence. At the classification stage, it presents a novel classifier that reaps the benefits of two different architectures: convolutional neural network and attention mechanism. The proposed framework is evaluated over the enhancer identification benchmark dataset where it outperforms the existing best-performing framework by 5%, and 9% in terms of accuracy and MCC. Similarly, when evaluated over the enhancer strength prediction benchmark dataset, it outperforms the existing best-performing framework by 4%, and 7% in terms of accuracy and MCC.

PREDAC-CNN: predicting antigenic clusters of seasonal influenza A viruses with convolutional neural network.

Vaccination stands as the most effective and economical strategy for prevention and control of influenza. The primary target of neutralizing antibodies is the surface antigen hemagglutinin (HA). However, ongoing mutations in the HA sequence result in antigenic drift. The success of a vaccine is contingent on its antigenic congruence with circulating strains. Thus, predicting antigenic variants and deducing antigenic clusters of influenza viruses are pivotal for recommendation of vaccine strains. The antigenicity of influenza A viruses is determined by the interplay of amino acids in the HA1 sequence. In this study, we exploit the ability of convolutional neural networks (CNNs) to extract spatial feature representations in the convolutional layers, which can discern interactions between amino acid sites. We introduce PREDAC-CNN, a model designed to track antigenic evolution of seasonal influenza A viruses. Accessible at http://predac-cnn.cloudna.cn, PREDAC-CNN formulates a spatially oriented representation of the HA1 sequence, optimized for the convolutional framework. It effectively probes interactions among amino acid sites in the HA1 sequence. Also, PREDAC-CNN focuses exclusively on physicochemical attributes crucial for the antigenicity of influenza viruses, thereby eliminating unnecessary amino acid embeddings. Together, PREDAC-CNN is adept at capturing interactions of amino acid sites within the HA1 sequence and examining the collective impact of point mutations on antigenic variation. Through 5-fold cross-validation and retrospective testing, PREDAC-CNN has shown superior performance in predicting antigenic variants compared to its counterparts. Additionally, PREDAC-CNN has been instrumental in identifying predominant antigenic clusters for A/H3N2 (1968-2023) and A/H1N1 (1977-2023) viruses, significantly aiding in vaccine strain recommendation.

Improved prediction of DNA and RNA binding proteins with deep learning models.

Nucleic acid-binding proteins (NABPs), including DNA-binding proteins (DBPs) and RNA-binding proteins (RBPs), play important roles in essential biological processes. To facilitate functional annotation and accurate prediction of different types of NABPs, many machine learning-based computational approaches have been developed. However, the datasets used for training and testing as well as the prediction scopes in these studies have limited their applications. In this paper, we developed new strategies to overcome these limitations by generating more accurate and robust datasets and developing deep learning-based methods including both hierarchical and multi-class approaches to predict the types of NABPs for any given protein. The deep learning models employ two layers of convolutional neural network and one layer of long short-term memory. Our approaches outperform existing DBP and RBP predictors with a balanced prediction between DBPs and RBPs, and are more practically useful in identifying novel NABPs. The multi-class approach greatly improves the prediction accuracy of DBPs and RBPs, especially for the DBPs with ~12% improvement. Moreover, we explored the prediction accuracy of single-stranded DNA binding proteins and their effect on the overall prediction accuracy of NABP predictions.

Enhancing genome-wide populus trait prediction through deep convolutional neural networks.

As a promising model, genome-based plant breeding has greatly promoted the improvement of agronomic traits. Traditional methods typically adopt linear regression models with clear assumptions, neither obtaining the linkage between phenotype and genotype nor providing good ideas for modification. Nonlinear models are well characterized in capturing complex nonadditive effects, filling this gap under traditional methods. Taking populus as the research object, this paper constructs a deep learning method, DCNGP, which can effectively predict the traits including 65 phenotypes. The method was trained on three datasets, and compared with other four classic models-Bayesian ridge regression (BRR), Elastic Net, support vector regression, and dualCNN. The results show that DCNGP has five typical advantages in performance: strong prediction ability on multiple experimental datasets; the incorporation of batch normalization layers and Early-Stopping technology enhancing the generalization capabilities and prediction stability on test data; learning potent features from the data and thus circumventing the tedious steps of manual production; the introduction of a Gaussian Noise layer enhancing predictive capabilities in the case of inherent uncertainties or perturbations; fewer hyperparameters aiding to reduce tuning time across datasets and improve auto-search efficiency. In this way, DCNGP shows powerful predictive ability from genotype to phenotype, which provide an important theoretical reference for building more robust populus breeding programs.

NanoDeep: a deep learning framework for nanopore adaptive sampling on microbial sequencing.

Nanopore sequencers can enrich or deplete the targeted DNA molecules in a library by reversing the voltage across individual nanopores. However, it requires substantial computational resources to achieve rapid operations in parallel at read-time sequencing. We present a deep learning framework, NanoDeep, to overcome these limitations by incorporating convolutional neural network and squeeze and excitation. We first showed that the raw squiggle derived from native DNA sequences determines the origin of microbial and human genomes. Then, we demonstrated that NanoDeep successfully classified bacterial reads from the pooled library with human sequence and showed enrichment for bacterial sequence compared with routine nanopore sequencing setting. Further, we showed that NanoDeep improves the sequencing efficiency and preserves the fidelity of bacterial genomes in the mock sample. In addition, NanoDeep performs well in the enrichment of metagenome sequences of gut samples, showing its potential applications in the enrichment of unknown microbiota. Our toolkit is available at https://github.com/lysovosyl/NanoDeep.

H2Opred: a robust and efficient hybrid deep learning model for predicting 2'-O-methylation sites in human RNA.

2'-O-methylation (2OM) is the most common post-transcriptional modification of RNA. It plays a crucial role in RNA splicing, RNA stability and innate immunity. Despite advances in high-throughput detection, the chemical stability of 2OM makes it difficult to detect and map in messenger RNA. Therefore, bioinformatics tools have been developed using machine learning (ML) algorithms to identify 2OM sites. These tools have made significant progress, but their performances remain unsatisfactory and need further improvement. In this study, we introduced H2Opred, a novel hybrid deep learning (HDL) model for accurately identifying 2OM sites in human RNA. Notably, this is the first application of HDL in developing four nucleotide-specific models [adenine (A2OM), cytosine (C2OM), guanine (G2OM) and uracil (U2OM)] as well as a generic model (N2OM). H2Opred incorporated both stacked 1D convolutional neural network (1D-CNN) blocks and stacked attention-based bidirectional gated recurrent unit (Bi-GRU-Att) blocks. 1D-CNN blocks learned effective feature representations from 14 conventional descriptors, while Bi-GRU-Att blocks learned feature representations from five natural language processing-based embeddings extracted from RNA sequences. H2Opred integrated these feature representations to make the final prediction. Rigorous cross-validation analysis demonstrated that H2Opred consistently outperforms conventional ML-based single-feature models on five different datasets. Moreover, the generic model of H2Opred demonstrated a remarkable performance on both training and testing datasets, significantly outperforming the existing predictor and other four nucleotide-specific H2Opred models. To enhance accessibility and usability, we have deployed a user-friendly web server for H2Opred, accessible at https://balalab-skku.org/H2Opred/. This platform will serve as an invaluable tool for accurately predicting 2OM sites within human RNA, thereby facilitating broader applications in relevant research endeavors.

A CNN-CBAM-BIGRU model for protein function prediction.

Understanding a protein's function based solely on its amino acid sequence is a crucial but intricate task in bioinformatics. Traditionally, this challenge has proven difficult. However, recent years have witnessed the rise of deep learning as a powerful tool, achieving significant success in protein function prediction. Their strength lies in their ability to automatically learn informative features from protein sequences, which can then be used to predict the protein's function. This study builds upon these advancements by proposing a novel model: CNN-CBAM+BiGRU. It incorporates a Convolutional Block Attention Module (CBAM) alongside BiGRUs. CBAM acts as a spotlight, guiding the CNN to focus on the most informative parts of the protein data, leading to more accurate feature extraction. BiGRUs, a type of Recurrent Neural Network (RNN), excel at capturing long-range dependencies within the protein sequence, which are essential for accurate function prediction. The proposed model integrates the strengths of both CNN-CBAM and BiGRU. This study's findings, validated through experimentation, showcase the effectiveness of this combined approach. For the human dataset, the suggested method outperforms the CNN-BIGRU+ATT model by +1.0 % for cellular components, +1.1 % for molecular functions, and +0.5 % for biological processes. For the yeast dataset, the suggested method outperforms the CNN-BIGRU+ATT model by +2.4 % for the cellular component, +1.2 % for molecular functions, and +0.6 % for biological processes.

DEEP-EP: Identification of epigenetic protein by ensemble residual convolutional neural network for drug discovery.

Epigenetic proteins (EP) play a role in the progression of a wide range of diseases, including autoimmune disorders, neurological disorders, and cancer. Recognizing their different functions has prompted researchers to investigate them as potential therapeutic targets and pharmacological targets. This paper proposes a novel deep learning-based model that accurately predicts EP. This study introduces a novel deep learning-based model that accurately predicts EP. Our approach entails generating two distinct datasets for training and evaluating the model. We then use three distinct strategies to transform protein sequences to numerical representations: Dipeptide Deviation from Expected Mean (DDE), Dipeptide Composition (DPC), and Group Amino Acid (GAAC). Following that, we train and compare the performance of four advanced deep learning models algorithms: Ensemble Residual Convolutional Neural Network (ERCNN), Generative Adversarial Network (GAN), Convolutional Neural Network (CNN), and Gated Recurrent Unit (GRU). The DDE encoding combined with the ERCNN model demonstrates the best performance on both datasets. This study demonstrates deep learning's potential for precisely predicting EP, which can considerably accelerate research and streamline drug discovery efforts. This analytical method has the potential to find new therapeutic targets and advance our understanding of EP activities in disease.

Deep learning based method for predicting DNA N6-methyladenosine sites.

DNA N6 methyladenine (6mA) plays an important role in many biological processes, and accurately identifying its sites helps one to understand its biological effects more comprehensively. Previous traditional experimental methods are very labor-intensive and traditional machine learning methods also seem to be somewhat insufficient as the database of 6mA methylation groups becomes progressively larger, so we propose a deep learning-based method called multi-scale convolutional model based on global response normalization (CG6mA) to solve the prediction problem of 6mA site. This method is tested with other methods on three different kinds of benchmark datasets, and the results show that our model can get more excellent prediction results.

Improving anti-cancer drug response prediction using multi-task learning on graph convolutional networks.

Predicting the therapeutic effect of anti-cancer drugs on tumors based on the characteristics of tumors and patients is one of the important contents of precision oncology. Existing computational methods regard the drug response prediction problem as a classification or regression task. However, few of them consider leveraging the relationship between the two tasks. In this work, we propose a Multi-task Interaction Graph Convolutional Network (MTIGCN) for anti-cancer drug response prediction. MTIGCN first utilizes an graph convolutional network-based model to produce embeddings for both cell lines and drugs. After that, the model employs multi-task learning to predict anti-cancer drug response, which involves training the model on three different tasks simultaneously: the main task of the drug sensitive or resistant classification task and the two auxiliary tasks of regression prediction and similarity network reconstruction. By sharing parameters and optimizing the losses of different tasks simultaneously, MTIGCN enhances the feature representation and reduces overfitting. The results of the experiments on two in vitro datasets demonstrated that MTIGCN outperformed seven state-of-the-art baseline methods. Moreover, the well-trained model on the in vitro dataset GDSC exhibited good performance when applied to predict drug responses in in vivo datasets PDX and TCGA. The case study confirmed the model's ability to discover unknown drug responses in cell lines.

Predicting lysine methylation sites using a convolutional neural network.

Protein lysine methylation is a particular type of post translational modification that plays an important role in both histone and non-histone function regulation in proteins. Deregulation caused by lysine methyltransferases has been identified as the cause of several diseases including cancer as well as both mental and developmental disorders. Identifying lysine methylation sites is a critical step in both early diagnosis and drug design. This study proposes a new Machine Learning method called CNN-Meth for predicting lysine methylation sites using a convolutional neural network (CNN). Our model is trained using evolutionary, structural, and physicochemical-based presentation along with binary encoding. Unlike previous studies, instead of extracting handcrafted features, we use CNN to automatically extract features from different presentations of amino acids to avoid information loss. Automated feature extraction from these representations of amino acids as well as CNN as a classifier have never been used for this problem. Our results demonstrate that CNN-Meth can significantly outperform previous methods for predicting methylation sites. It achieves 96.0%, 85.1%, 96.4%, and 0.65 in terms of Accuracy, Sensitivity, Specificity, and Matthew's Correlation Coefficient (MCC), respectively. CNN-Meth and its source code are publicly available at https://github.com/MLBC-lab/CNN-Meth.

Multiclass classification of motor imagery tasks based on multi-branch convolutional neural network and temporal convolutional network model.

Motor imagery (MI) is a cognitive process wherein an individual mentally rehearses a specific movement without physically executing it. Recently, MI-based brain-computer interface (BCI) has attracted widespread attention. However, accurate decoding of MI and understanding of neural mechanisms still face huge challenges. These seriously hinder the clinical application and development of BCI systems based on MI. Thus, it is very necessary to develop new methods to decode MI tasks. In this work, we propose a multi-branch convolutional neural network (MBCNN) with a temporal convolutional network (TCN), an end-to-end deep learning framework to decode multi-class MI tasks. We first used MBCNN to capture the MI electroencephalography signals information on temporal and spectral domains through different convolutional kernels. Then, we introduce TCN to extract more discriminative features. The within-subject cross-session strategy is used to validate the classification performance on the dataset of BCI Competition IV-2a. The results showed that we achieved 75.08% average accuracy for 4-class MI task classification, outperforming several state-of-the-art approaches. The proposed MBCNN-TCN-Net framework successfully captures discriminative features and decodes MI tasks effectively, improving the performance of MI-BCIs. Our findings could provide significant potential for improving the clinical application and development of MI-based BCI systems.

Utilizing graph convolutional networks for identification of mild cognitive impairment from single modal fMRI data: a multiconnection pattern combination approach.

Mild cognitive impairment plays a crucial role in predicting the early progression of Alzheimer's disease, and it can be used as an important indicator of the disease progression. Currently, numerous studies have focused on utilizing the functional brain network as a novel biomarker for mild cognitive impairment diagnosis. In this context, we employed a graph convolutional neural network to automatically extract functional brain network features, eliminating the need for manual feature extraction, to improve the mild cognitive impairment diagnosis performance. However, previous graph convolutional neural network approaches have primarily concentrated on single modes of brain connectivity, leading to a failure to leverage the potential complementary information offered by diverse connectivity patterns and limiting their efficacy. To address this limitation, we introduce a novel method called the graph convolutional neural network with multimodel connectivity, which integrates multimode connectivity for the identification of mild cognitive impairment using fMRI data and evaluates the graph convolutional neural network with multimodel connectivity approach through a mild cognitive impairment diagnostic task on the Alzheimer's Disease Neuroimaging Initiative dataset. Overall, our experimental results show the superiority of the proposed graph convolutional neural network with multimodel connectivity approach, achieving an accuracy rate of 92.2% and an area under the Receiver Operating Characteristic (ROC) curve of 0.988.

SEnSCA: Identifying possible ligand-receptor interactions and its application in cell-cell communication inference.

Multicellular organisms have dense affinity with the coordination of cellular activities, which severely depend on communication across diverse cell types. Cell-cell communication (CCC) is often mediated via ligand-receptor interactions (LRIs). Existing CCC inference methods are limited to known LRIs. To address this problem, we developed a comprehensive CCC analysis tool SEnSCA by integrating single cell RNA sequencing and proteome data. SEnSCA mainly contains potential LRI acquisition and CCC strength evaluation. For acquiring potential LRIs, it first extracts LRI features and reduces the feature dimension, subsequently constructs negative LRI samples through K-means clustering, finally acquires potential LRIs based on Stacking ensemble comprising support vector machine, 1D-convolutional neural networks and multi-head attention mechanism. During CCC strength evaluation, SEnSCA conducts LRI filtering and then infers CCC by combining the three-point estimation approach and single cell RNA sequencing data. SEnSCA computed better precision, recall, accuracy, F1 score, AUC and AUPR under most of conditions when predicting possible LRIs. To better illustrate the inferred CCC network, SEnSCA provided three visualization options: heatmap, bubble diagram and network diagram. Its application on human melanoma tissue demonstrated its reliability in CCC detection. In summary, SEnSCA offers a useful CCC inference tool and is freely available at https://github.com/plhhnu/SEnSCA.

A machine-learning tool to identify bistable states from calcium imaging data.

Mapping neuronal activation using calcium imaging in vivo during behavioural tasks has advanced our understanding of nervous system function. In almost all of these studies, calcium imaging is used to infer spike probabilities because action potentials activate voltage-gated calcium channels and increase intracellular calcium levels. However, neurons not only fire action potentials, but also convey information via intrinsic dynamics such as by generating bistable membrane potential states. Although a number of tools for spike inference have been developed and are currently being used, no tool exists for converting calcium imaging signals to maps of cellular state in bistable neurons. Purkinje neurons in the larval zebrafish cerebellum exhibit membrane potential bistability, firing either tonically or in bursts. Several studies have implicated the role of a population code in cerebellar function, with bistability adding an extra layer of complexity to this code. In the present study, we develop a tool, CaMLSort, which uses convolutional recurrent neural networks to classify calcium imaging traces as arising from either tonic or bursting cells. We validate this classifier using a number of different methods and find that it performs well on simulated event rasters as well as real biological data that it had not previously seen. Moreover, we find that CaMLsort generalizes to other bistable neurons, such as dopaminergic neurons in the ventral tegmental area of mice. Thus, this tool offers a new way of analysing calcium imaging data from bistable neurons to understand how they participate in network computation and natural behaviours. KEY POINTS: Calcium imaging, compriising the gold standard of inferring neuronal activity, does not report cellular state in neurons that are bistable, such as Purkinje neurons in the cerebellum of larval zebrafish. We model the relationship between Purkinje neuron electrical activity and its corresponding calcium signal to compile a dataset of state-labelled simulated calcium signals. We apply machine-learning methods to this dataset to develop a tool that can classify the state of a Purkinje neuron using only its calcium signal, which works well on real data even though it was trained only on simulated data. CaMLsort (Calcium imaging and Machine Learning based tool to sort intracellular state) also generalizes well to bistable neurons in a different brain region (ventral tegmental area) in a different model organism (mouse). This tool can facilitate our understanding of how these neurons carry out their functions in a circuit.

Analysis and prediction of interactions between transmembrane and non-transmembrane proteins.

BACKGROUND: Most of the important biological mechanisms and functions of transmembrane proteins (TMPs) are realized through their interactions with non-transmembrane proteins(nonTMPs). The interactions between TMPs and nonTMPs in cells play vital roles in intracellular signaling, energy metabolism, investigating membrane-crossing mechanisms, correlations between disease and drugs. RESULTS: Despite the importance of TMP-nonTMP interactions, the study of them remains in the wet experimental stage, lacking specific and comprehensive studies in the field of bioinformatics. To fill this gap, we performed a comprehensive statistical analysis of known TMP-nonTMP interactions and constructed a deep learning-based predictor to identify potential interactions. The statistical analysis describes known TMP-nonTMP interactions from various perspectives, such as distributions of species and protein families, enrichment of GO and KEGG pathways, as well as hub proteins and subnetwork modules in the PPI network. The predictor implemented by an end-to-end deep learning model can identify potential interactions from protein primary sequence information. The experimental results over the independent validation demonstrated considerable prediction performance with an MCC of 0.541. CONCLUSIONS: To our knowledge, we were the first to focus on TMP-nonTMP interactions. We comprehensively analyzed them using bioinformatics methods and predicted them via deep learning-based solely on their sequence. This research completes a key link in the protein network, benefits the understanding of protein functions, and helps in pathogenesis studies of diseases and associated drug development.

Tinnitus classification based on resting-state functional connectivity using a convolutional neural network architecture.

OBJECTIVES: Many studies have investigated aberrant functional connectivity (FC) using resting-state functional MRI (rs-fMRI) in subjective tinnitus patients. However, no studies have verified the efficacy of resting-state FC as a diagnostic imaging marker. We established a convolutional neural network (CNN) model based on rs-fMRI FC to distinguish tinnitus patients from healthy controls, providing guidance and fast diagnostic tools for the clinical diagnosis of subjective tinnitus. METHODS: A CNN architecture was trained on rs-fMRI data from 100 tinnitus patients and 100 healthy controls using an asymmetric convolutional layer. Additionally, a traditional machine learning model and a transfer learning model were included for comparison with the CNN, and each of the three models was tested on three different brain atlases. RESULTS: Of the three models, the CNN model outperformed the other two models with the highest area under the curve, especially on the Dos_160 atlas (AUC = 0.944). Meanwhile, the model with the best classification performance highlights the crucial role of the default mode network, salience network, and sensorimotor network in distinguishing between normal controls and patients with subjective tinnitus. CONCLUSION: Our CNN model could appropriately tackle the diagnosis of tinnitus patients using rs-fMRI and confirmed the diagnostic value of FC as measured by rs-fMRI.

A 3D convolutional neural network to classify subjects as Alzheimer's disease, frontotemporal dementia or healthy controls using brain 18F-FDG PET.

With the arrival of disease-modifying drugs, neurodegenerative diseases will require an accurate diagnosis for optimal treatment. Convolutional neural networks are powerful deep learning techniques that can provide great help to physicians in image analysis. The purpose of this study is to introduce and validate a 3D neural network for classification of Alzheimer's disease (AD), frontotemporal dementia (FTD) or cognitively normal (CN) subjects based on brain glucose metabolism. Retrospective [18F]-FDG-PET scans of 199 CE, 192 FTD and 200 CN subjects were collected from our local database, Alzheimer's disease and frontotemporal lobar degeneration neuroimaging initiatives. Training and test sets were created using randomization on a 90 %-10 % basis, and training of a 3D VGG16-like neural network was performed using data augmentation and cross-validation. Performance was compared to clinical interpretation by three specialists in the independent test set. Regions determining classification were identified in an occlusion experiment and Gradient-weighted Class Activation Mapping. Test set subjects were age- and sex-matched across categories. The model achieved an overall 89.8 % accuracy in predicting the class of test scans. Areas under the ROC curves were 93.3 % for AD, 95.3 % for FTD, and 99.9 % for CN. The physicians' consensus showed a 69.5 % accuracy, and there was substantial agreement between them (kappa = 0.61, 95 % CI: 0.49-0.73). To our knowledge, this is the first study to introduce a deep learning model able to discriminate AD and FTD based on [18F]-FDG PET scans, and to isolate CN subjects with excellent accuracy. These initial results are promising and hint at the potential for generalization to data from other centers.

Brain age prediction using interpretable multi-feature-based convolutional neural network in mild traumatic brain injury.

BACKGROUND: Convolutional neural network (CNN) can capture the structural features changes of brain aging based on MRI, thus predict brain age in healthy individuals accurately. However, most studies use single feature to predict brain age in healthy individuals, ignoring adding information from multiple sources and the changes in brain aging patterns after mild traumatic brain injury (mTBI) were still unclear. METHODS: Here, we leveraged the structural data from a large, heterogeneous dataset (N = 1464) to implement an interpretable 3D combined CNN model for brain-age prediction. In addition, we also built an atlas-based occlusion analysis scheme with a fine-grained human Brainnetome Atlas to reveal the age-sstratified contributed brain regions for brain-age prediction in healthy controls (HCs) and mTBI patients. The correlations between brain predicted age gaps (brain-PAG) following mTBI and individual's cognitive impairment, as well as the level of plasma neurofilament light were also examined. RESULTS: Our model utilized multiple 3D features derived from T1w data as inputs, and reduced the mean absolute error (MAE) of age prediction to 3.08 years and improved Pearson's r to 0.97 on 154 HCs. The strong generalizability of our model was also validated across different centers. Regions contributing the most significantly to brain age prediction were the caudate and thalamus for HCs and patients with mTBI, and the contributive regions were mostly located in the subcortical areas throughout the adult lifespan. The left hemisphere was confirmed to contribute more in brain age prediction throughout the adult lifespan. Our research showed that brain-PAG in mTBI patients was significantly higher than that in HCs in both acute and chronic phases. The increased brain-PAG in mTBI patients was also highly correlated with cognitive impairment and a higher level of plasma neurofilament light, a marker of neurodegeneration. The higher brain-PAG and its correlation with severe cognitive impairment showed a longitudinal and persistent nature in patients with follow-up examinations. CONCLUSION: We proposed an interpretable deep learning framework on a relatively large dataset to accurately predict brain age in both healthy individuals and mTBI patients. The interpretable analysis revealed that the caudate and thalamus became the most contributive role across the adult lifespan in both HCs and patients with mTBI. The left hemisphere contributed significantly to brain age prediction may enlighten us to be concerned about the lateralization of brain abnormality in neurological diseases in the future. The proposed interpretable deep learning framework might also provide hope for testing the performance of related drugs and treatments in the future.