Comput Methods Programs Biomed. 2024 Sep 30;257:108448. doi: 10.1016/j.cmpb.2024.108448. Online ahead of print.

ABSTRACT

BACKGROUND AND OBJECTIVE: Despite recent performance advancements, deep learning models are not yet adopted in clinical practice on a wide scale. The intrinsic intransparency of such systems is commonly cited as one major reason for this reluctance. This has motivated methods that aim to provide explanations of model functioning. Known limitations of feature-based explanations have led to an increased interest in concept-based interpretability. Testing with Concept Activation Vectors (TCAV) employs human-understandable, abstract concepts to explain model behavior. The method has previously been applied to the medical domain in the context of electronic health records, retinal fundus images and magnetic resonance imaging.

METHODS: We explore the usage of TCAV for building interpretable models on physiological time series, using an example of abnormality detection in electroencephalography (EEG). For this purpose, we adopt the XceptionTime model, which is suitable for multi-channel physiological data of variable sizes. The model provides state-of-the-art performance on raw EEG data and is publically available. We propose and test several ideas regarding concept definition through metadata mining, using additional labeled EEG data and extracting interpretable signal characteristics in the form of frequencies. By including our own hospital data with analog labeling, we further evaluate the robustness of our approach.

RESULTS: The tested concepts show a TCAV score distribution that is in line with the clinical expectations, i.e. concepts known to have strong links with EEG pathologies (such as epileptiform discharges) received higher scores than the neutral concepts (e.g. sex). The scores were consistent across the applied concept generation strategies.

CONCLUSIONS: TCAV has the potential to improve interpretability of deep learning applied to multi-channel signals as well as to detect possible biases in the data. Still, further work on developing the strategies for concept definition and validation on clinical physiological time series is needed to better understand how to extract clinically relevant information from the concept sensitivity scores.

PMID:39395304 | DOI:10.1016/j.cmpb.2024.108448

Comput Methods Programs Biomed. 2024 Oct 5;257:108451. doi: 10.1016/j.cmpb.2024.108451. Online ahead of print.

ABSTRACT

BACKGROUND AND OBJECTIVE: Fetal brain tissue segmentation provides foundational support for comprehensively understanding the neurodevelopment of normal and congenital disease-affected fetuses. Manual labeling is very time-consuming, and automated segmentation methods can greatly improve the efficiency of doctors. At the same time, fetal brain tissue undergoes various changes throughout the pregnancy, leading to a continuous change in tissue contrast, which greatly increases the difficulty of training segmentation methods. This study aims to develop an automated segmentation model that can efficiently and accurately segment fetal brain tissue, improving the workflow for medical professionals.

METHODS: We propose a novel deep learning-based segmentation model that incorporates three innovative components: Firstly, a new Dual Dilated Attention Block (DDAB) is proposed in the encoder part to enhance the feature extraction of local spatial and structural contextual information. Secondly, a Multi-scale Deformable Transformer (MSDT) is integrated into the bottleneck to improve the feature extraction of global information on local spatial and structural contextual information. Thirdly, we use a novel block based on Graph Convolution Attention (GCAB) in the decoder, which effectively enhances the features at the decoder.The code is available at https://github.com/unicoco7/MG-Net/.

RESULTS: We trained and tested on the FeTA 2021 and FeTA 2022 datasets, and evaluated using seven popular metrics, including Dice, IoU, MAE, BoundaryF, PRE, SEN, and SPE. Compared to the current state-of-the-art 3D segmentation models such as nnFormer, SwinUNETR, and 3DUX-net, our proposed method has surpassed all of them in metrics like Dice, IoU, and MAE. Specifically, on the FeTA 2021 dataset, our model achieved a Dice of 0.8666, an IoU of 0.7646, and an MAE of 0.0027; on the FeTA 2022 dataset, it achieved a Dice of 0.8552, an IoU of 0.7470, and an MAE of 0.0005.

CONCLUSION: In this paper, we propose a model for three-dimensional fetal brain tissue segmentation based on multi-scale feature fusion and graph convolution attention mechanism, and conduct experimental evaluation on the FeTA 2021 and FeTA 2022 datasets. Understanding the boundaries of fetal brain tissue is crucial for doctors' diagnosis, so the proposed model is expected to improve the speed and accuracy of doctors' diagnoses.

PMID:39395303 | DOI:10.1016/j.cmpb.2024.108451

Pathol Res Pract. 2024 Oct 5;263:155644. doi: 10.1016/j.prp.2024.155644. Online ahead of print.

ABSTRACT

Breast cancer (BC) is the most frequently occurring cancer disease observed in women after lung cancer. Out of different stages, invasive ductal BC causes maximum deaths in women. In this work, three deep learning (DL) models such as Vision Transformer (ViT), Convmixer, and Visual Geometry Group-19 (VGG-19) are implemented for the detection and classification of different breast cancer tumors with the help of Breast cancer histopathological (Break His) image database. The performance of each model is evaluated using an 80:20 training scheme and measured in terms of accuracy, precision, recall, loss, F1-score, and area under the curve (AUC). From the simulation result, ViT showed the best performance for binary classification of breast cancer tumors with accuracy, precision, recall, and F1-score of 99.89 %, 98.29 %, 98.29 %, and 98.29 %, respectively. Also, ViT showed the best performance in terms of accuracy (98.21 %), average Precision (89.84 %), recall (89.97 %), and F1-score (88.75) for eight class classifications. Moreover, we have also ensemble the ViT-Convmixer model and observed that the performance of the ensemble model is reduced as compared to the ViT model. We have also compared the performance of the proposed best model with other existing models reported by several research groups. The study will help find suitable models that will increase accuracy in early diagnoses of BC. We hope the study will also help to minimize human errors in the early diagnosis of this fatal disease and administer appropriate treatment. The proposed model may also be implemented for the detection of other diseases with improved accuracy.

PMID:39395299 | DOI:10.1016/j.prp.2024.155644

Med Image Anal. 2024 Oct 9;99:103365. doi: 10.1016/j.media.2024.103365. Online ahead of print.

ABSTRACT

In the last decades, many publicly available large fundus image datasets have been collected for diabetic retinopathy, glaucoma, and age-related macular degeneration, and a few other frequent pathologies. These publicly available datasets were used to develop a computer-aided disease diagnosis system by training deep learning models to detect these frequent pathologies. One challenge limiting the adoption of a such system by the ophthalmologist is, computer-aided disease diagnosis system ignores sight-threatening rare pathologies such as central retinal artery occlusion or anterior ischemic optic neuropathy and others that ophthalmologists currently detect. Aiming to advance the state-of-the-art in automatic ocular disease classification of frequent diseases along with the rare pathologies, a grand challenge on "Retinal Image Analysis for multi-Disease Detection" was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI - 2021). This paper, reports the challenge organization, dataset, top-performing participants solutions, evaluation measures, and results based on a new "Retinal Fundus Multi-disease Image Dataset" (RFMiD). There were two principal sub-challenges: disease screening (i.e. presence versus absence of pathology - a binary classification problem) and disease/pathology classification (a 28-class multi-label classification problem). It received a positive response from the scientific community with 74 submissions by individuals/teams that effectively entered in this challenge. The top-performing methodologies utilized a blend of data-preprocessing, data augmentation, pre-trained model, and model ensembling. This multi-disease (frequent and rare pathologies) detection will enable the development of generalizable models for screening the retina, unlike the previous efforts that focused on the detection of specific diseases.

PMID:39395210 | DOI:10.1016/j.media.2024.103365

Med Phys. 2024 Oct 12. doi: 10.1002/mp.17441. Online ahead of print.

ABSTRACT

BACKGROUND: Although the uterus, bladder, and rectum are distinct organs, their muscular fasciae are often interconnected. Clinical experience suggests that they may share common risk factors and associations. When one organ experiences prolapse, it can potentially affect the neighboring organs. However, the current assessment of disease severity still relies on manual measurements, which can yield varying results depending on the physician, thereby leading to diagnostic inaccuracies.

PURPOSE: This study aims to develop a multilabel grading model based on deep learning to classify the degree of prolapse of three organs in the female pelvis using stress magnetic resonance imaging (MRI) and provide interpretable result analysis.

METHODS: We utilized sagittal MRI sequences taken at rest and during maximum Valsalva maneuver from 662 subjects. The training set included 464 subjects, the validation set included 98 subjects, and the test set included 100 subjects (training set n = 464, validation set n = 98, test set n = 100). We designed a feature extraction module specifically for pelvic floor MRI using the vision transformer architecture and employed label masking training strategy and pre-training methods to enhance model convergence. The grading results were evaluated using Precision, Kappa, Recall, and Area Under the Curve (AUC). To validate the effectiveness of the model, the designed model was compared with classic grading methods. Finally, we provided interpretability charts illustrating the model's operational principles on the grading task.

RESULTS: In terms of POP grading detection, the model achieved an average Precision, Kappa coefficient, Recall, and AUC of 0.86, 0.77, 0.76, and 0.86, respectively. Compared to existing studies, our model achieved the highest performance metrics. The average time taken to diagnose a patient was 0.38 s.

CONCLUSIONS: The proposed model achieved detection accuracy that is comparable to or even exceeds that of physicians, demonstrating the effectiveness of the vision transformer architecture and label masking training strategy for assisting in the grading of POP under static and maximum Valsalva conditions. This offers a promising option for computer-aided diagnosis and treatment planning of POP.

PMID:39395206 | DOI:10.1002/mp.17441

Angew Chem Int Ed Engl. 2024 Oct 12:e202414705. doi: 10.1002/anie.202414705. Online ahead of print.

ABSTRACT

Deep learning tools for enzyme design are rapidly emerging, and there is a critical need to evaluate their effectiveness in engineering workflows. Here we show that the deep learning-based tool ProteinMPNN can be used to redesign Fe(II)/αKG superfamily enzymes for greater stability, solubility, and expression while retaining both native activity and industrially-relevant non-native functions. This superfamily has diverse catalytic functions and could provide a rich new source of biocatalysts for synthesis and industrial processes. Through systematic comparisons of directed evolution trajectories for a non-native, remote C(sp3)-H hydroxylation reaction, we demonstrate that the stabilized redesign can be evolved more efficiently than the wild-type enzyme. After three rounds of directed evolution, we obtained a 6-fold activity increase from the wild-type parent and an 80-fold increase from the stabilized variant. To generate the initial stabilized variant, we identified multiple structural and sequence constraints to preserve catalytic function. We applied these criteria to produce stabilized, catalytically active variants of a second Fe(II)/αKG enzyme, suggesting that the approach is generalizable to additional members of the Fe(II)/αKG superfamily. ProteinMPNN is user-friendly and widely-accessible, and our results provide a framework for the routine implementation of deep learning-based protein stabilization tools in directed evolution workflows for novel biocatalysts.

PMID:39394803 | DOI:10.1002/anie.202414705

Biomed Phys Eng Express. 2024 Oct 11;10(6). doi: 10.1088/2057-1976/ad81ff.

ABSTRACT

The integration of artificial intelligence (AI) into lung cancer management offers immense potential to revolutionize diagnostic and treatment strategies. The aim is to develop a resilient AI framework capable of two critical tasks: firstly, achieving accurate and automated segmentation of lung tumors and secondly, facilitating the T classification of lung cancer according to the ninth edition of TNM staging 2024 based on PET/CT imaging. This study presents a robust AI framework for the automated segmentation of lung tumors and T classification of lung cancer using PET/CT imaging. The database includes axial DICOM CT and18FDG-PET/CT images. A modified ResNet-50 model was employed for segmentation, achieving high precision and specificity. Reconstructed 3D models of segmented slices enhance tumor boundary visualization, which is essential for treatment planning. The Pulmonary Toolkit facilitated lobe segmentation, providing critical diagnostic insights. Additionally, the segmented images were used as input for the T classification using a CNN ResNet-50 model. Our classification model demonstrated excellent performance, particularly for T1a, T2a, T2b, T3 and T4 tumors, with high precision, F1 scores, and specificity. The T stage is particularly relevant in lung cancer as it determines treatment approaches (surgery, chemotherapy and radiation therapy or supportive care) and prognosis assessment. In fact, for Tis-T2, each increase of one centimeter in tumor size results in a worse prognosis. For locally advanced tumors (T3-T4) and regardless of size, the prognosis is poorer. This AI framework marks a significant advancement in the automation of lung cancer diagnosis and staging, promising improved patient outcomes.

PMID:39394688 | DOI:10.1088/2057-1976/ad81ff

Neurosurg Rev. 2024 Oct 12;47(1):788. doi: 10.1007/s10143-024-03028-1.

ABSTRACT

Central nervous system tumors, classified by the WHO into four grades based on their aggressiveness, present significant challenges in treatment, particularly low-grade gliomas (LGGs) which, despite their slower growth, can progress to high-grade gliomas. Lucca B. Palavani and colleagues evaluated the efficacy and safety of SBT I-125 brachytherapy for LGMs in a systematic review and meta-analysis of 20 studies involving 988 patients. The analysis revealed an overall complication rate of 10%, with headaches and cyst formation being the most frequent issues. The five-year progression-free survival (PFS) rate was 66%, while the ten-year PFS rate was 30%, and the rate of malignant transformation was 26%. The mortality rate was 33%. Despite these findings, significant limitations were noted, including data insufficiencies, study heterogeneity, lack of randomized controlled trials, and potential publication bias. Inconsistencies in follow-up durations further hindered the evaluation of long-term efficacy and safety. Recent advancements in automated tumor assessment, such as Cheng et al.'s deep learning-based pipeline, are revolutionizing glioma management by enhancing the accuracy and consistency of tumor volume and RANO measurements. These innovations facilitate improved glioma grading, genetic mutation prediction, surgical planning, real-time intraoperative guidance, and histopathological analysis. Integrating such advanced tools into clinical practice can significantly enhance the precision and efficiency of glioma management. In conclusion, while SBT I-125 brachytherapy shows promise, concerns regarding safety and efficacy underscore the need for further research with standardized methodologies. Incorporating advanced automated assessment tools could improve treatment evaluation and patient outcomes.

PMID:39394531 | DOI:10.1007/s10143-024-03028-1

Hypertens Res. 2024 Oct 12. doi: 10.1038/s41440-024-01938-7. Online ahead of print.

ABSTRACT

Arterial hypertension is a major risk factor for cardiovascular diseases. While cardiac ultrasound is a typical way to diagnose hypertension-mediated heart change, it often fails to detect early subtle structural changes. Electrocardiogram(ECG) represents electrical activity of heart muscle, affected by the changes in heart's structure. It is crucial to explore whether ECG can capture slight signals of hypertension-mediated heart change. However, reading ECG records is complex and some signals are too subtle to be captured by cardiologist's visual inspection. In this study, we designed a deep learning model to predict hypertension on ECG signals and then to identify hypertension-associated ECG segments. From The First Affiliated Hospital of Xiamen University, we collected 210,120 10-s 12-lead ECGs using the FX-8322 manufactured by FUKUDA and 812 ECGs using the RAGE-12 manufactured by NALONG. We proposed a deep learning framework, including MML-Net, a multi-branch, multi-scale LSTM neural network to evaluate the potential of ECG signals to detect hypertension, and ECG-XAI, an ECG-oriented wave-alignment AI explanation pipeline to identify hypertension-associated ECG segments. MML-Net achieved an 82% recall and an 87% precision in the testing, and an 80% recall and an 82% precision in the independent testing. In contrast, experienced clinical cardiologists typically attain recall rates ranging from 30 to 50% by visual inspection. The experiments demonstrate that ECG signals are sensitive to slight changes in heart structure caused by hypertension. ECG-XAI detects that R-wave and P-wave are the hypertension-associated ECG segments. The proposed framework has the potential to facilitate early diagnosis of heart change.

PMID:39394520 | DOI:10.1038/s41440-024-01938-7

Nat Comput Sci. 2024 Oct 11. doi: 10.1038/s43588-024-00703-7. Online ahead of print.

ABSTRACT

Anomalous diffusion plays a crucial rule in understanding molecular-level dynamics by offering valuable insights into molecular interactions, mobility states and the physical properties of systems across both biological and materials sciences. Deep-learning techniques have recently outperformed conventional statistical methods in anomalous diffusion recognition. However, deep-learning networks are typically trained by data with limited distribution, which inevitably fail to recognize unknown diffusion models and misinterpret dynamics when confronted with out-of-distribution (OOD) scenarios. In this work, we present a general framework for evaluating deep-learning-based OOD dynamics-detection methods. We further develop a baseline approach that achieves robust OOD dynamics detection as well as accurate recognition of in-distribution anomalous diffusion. We demonstrate that this method enables a reliable characterization of complex behaviors across a wide range of experimentally diverse systems, including nicotinic acetylcholine receptors in membranes, fluorescent beads in dextran solutions and silver nanoparticles undergoing active endocytosis.

PMID:39394501 | DOI:10.1038/s43588-024-00703-7

Nat Comput Sci. 2024 Oct 11. doi: 10.1038/s43588-024-00705-5. Online ahead of print.

NO ABSTRACT

PMID:39394500 | DOI:10.1038/s43588-024-00705-5

Nat Biotechnol. 2024 Oct 11. doi: 10.1038/s41587-024-02414-w. Online ahead of print.

ABSTRACT

A systematic evaluation of how model architectures and training strategies impact genomics model performance is needed. To address this gap, we held a DREAM Challenge where competitors trained models on a dataset of millions of random promoter DNA sequences and corresponding expression levels, experimentally determined in yeast. For a robust evaluation of the models, we designed a comprehensive suite of benchmarks encompassing various sequence types. All top-performing models used neural networks but diverged in architectures and training strategies. To dissect how architectural and training choices impact performance, we developed the Prix Fixe framework to divide models into modular building blocks. We tested all possible combinations for the top three models, further improving their performance. The DREAM Challenge models not only achieved state-of-the-art results on our comprehensive yeast dataset but also consistently surpassed existing benchmarks on Drosophila and human genomic datasets, demonstrating the progress that can be driven by gold-standard genomics datasets.

PMID:39394483 | DOI:10.1038/s41587-024-02414-w

EJNMMI Phys. 2024 Oct 12;11(1):84. doi: 10.1186/s40658-024-00686-4.

ABSTRACT

BACKGROUND: Accurate attenuation correction (AC) is vital in nuclear medicine, particularly for quantitative single-photon emission computed tomography/computed tomography (SPECT/CT) imaging. This study aimed to establish a CT-free quantification technology in kidney SPECT imaging using deep learning to generate synthetic attenuation maps (μ-maps) from SPECT data, thereby reducing radiation exposure and eliminating the need for CT scans.

RESULTS: A dataset of 1000 Tc-99m DTPA SPECT/CT scans was analyzed for training (n = 800), validation (n = 100), and testing (n = 100) using a modified 3D U-Net for deep learning. The study investigated the use of primary emission and scattering SPECT data, normalization methods, loss function optimization, and up-sampling techniques for optimal μ-map generation. The problem of checkerboard artifacts, unique to μ-map generation from SPECT signals, and the effects of iodine contrast media were evaluated. The addition of scattering SPECT to primary emission SPECT imaging, logarithmic maximum normalization, the combination of absolute difference loss (L1) and three times the absolute gradient difference loss (3 × LGDL), and the nearest-neighbor interpolation significantly enhanced AI performance in μ-map generation (p < 0.00001). Checkerboard artifacts were effectively eliminated using the nearest-neighbor interpolation technique. The developed AI algorithm produced μ-maps neutral to the presence of iodine contrast and showed negligible contrast effects on quantitative SPECT measurement, such as glomerular filtration rate (GFR). The potential reduction in radiation exposure by transitioning to AI-based CT-free SPECT imaging ranges from 45.3 to 78.8%.

CONCLUSION: The study successfully developed and optimized a deep learning algorithm for generating synthetic μ-maps in kidney SPECT images, demonstrating the potential to transition from conventional SPECT/CT to CT-free SPECT imaging for GFR measurement. This advancement represents a significant step towards enhancing patient safety and efficiency in nuclear medicine.

PMID:39394395 | DOI:10.1186/s40658-024-00686-4

Sci Rep. 2024 Oct 11;14(1):23830. doi: 10.1038/s41598-024-74320-9.

ABSTRACT

In this paper, we propose an efficient target detection algorithm, ASF-Sim-YOLO, to address issues encountered in escalator step defect detection, such as an excessive number of parameters in the detection network model, poor adaptability, and difficulties in real-time processing of video streams. Firstly, to address the characteristics of escalator step defects, we designed the ASF-Sim-P2 structure to improve the detection accuracy of small targets, such as step defects. Additionally, we incorporated the SimAM (Similarity-based Attention Mechanism) by combining SimAM with SPPF (Spatial Pyramid Pooling-Fast) to enhance the model's ability to capture key information by assigning importance weights to each pixel. Furthermore, to address the challenge posed by the small size of step defects, we replaced the traditional CIoU (Complete-Intersection-over-Union) loss function with NWD (Normalized Wasserstein Distance), which alleviated the problem of defect missing. Finally, to meet the deployment requirements of mobile devices, we performed channel pruning on the model. The experimental results showed that the improved ASF-Sim-YOLO model achieved an average accuracy (mAP50) of 96.8% on the test data set, which was a 22.1% improvement in accuracy compared to the baseline model. Meanwhile, the computational complexity (in GFLOPS) of the model was reduced to a quarter of that of the baseline model, while the frame rate (FPS) was improved to 575.1. Compared with YOLOv3-tiny, YOLOv5s, YOLOv8s, Faster-RCNN, TOOD, RTMDET and other deep learning-based target recognition algorithms, ASF-Sim-YOLO has better detection accuracy and real-time processing capability. These results demonstrate that ASF-Sim-YOLO effectively balances lightweight design and performance improvement, making it highly suitable for real-time detection of step defects, which can meet the demands of escalator inspection operations.

PMID:39394361 | DOI:10.1038/s41598-024-74320-9

Sci Data. 2024 Oct 11;11(1):1122. doi: 10.1038/s41597-024-03948-z.

ABSTRACT

Leaf Area Index (LAI) is a critical parameter in terrestrial ecosystems, with high spatial resolution data being extensively utilized in various research studies. However, LAI data under future scenarios are typically only available at 1° or coarser spatial resolutions. In this study, we generated a dataset of 0.05° LAI (F0.05D-LAI) from 1983-2100 in a high spatial resolution using the LAI Downscaling Network (LAIDN) model driven by inputs including air temperature, relative humidity, precipitation, and topography data. The dataset spans the historical period (1983-2014) and future scenarios (2015-2100, including SSP-126, SSP-245, SSP-370, and SSP-585) with a monthly interval. It achieves high accuracy (R² = 0.887, RMSE = 0.340) and captures fine spatial details across various climate zones and terrain types, indicating a slightly greening trend under future scenarios. F0.05D-LAI is the first high-resolution LAI dataset and reveals the potential vegetation variation under future scenarios in China, which benefits vegetation studies and model development in earth and environmental sciences across present and future periods.

PMID:39394222 | DOI:10.1038/s41597-024-03948-z

Nat Commun. 2024 Oct 11;15(1):8798. doi: 10.1038/s41467-024-53114-7.

ABSTRACT

Transposable elements (TEs) are crucial for genetic diversity and gene regulation. Current single-cell quantification methods often align multi-mapping reads to either 'best-mapped' or 'random-mapped' locations and categorize them at the subfamily levels, overlooking the biological necessity for accurate, locus-specific TE quantification. Moreover, these existing methods are primarily designed for and focused on transcriptomics data, which restricts their adaptability to single-cell data of other modalities. To address these challenges, here we introduce MATES, a deep-learning approach that accurately allocates multi-mapping reads to specific loci of TEs, utilizing context from adjacent read alignments flanking the TE locus. When applied to diverse single-cell omics datasets, MATES shows improved performance over existing methods, enhancing the accuracy of TE quantification and aiding in the identification of marker TEs for identified cell populations. This development facilitates the exploration of single-cell heterogeneity and gene regulation through the lens of TEs, offering an effective transposon quantification tool for the single-cell genomics community.

PMID:39394211 | DOI:10.1038/s41467-024-53114-7

Nat Commun. 2024 Oct 11;15(1):8808. doi: 10.1038/s41467-024-53116-5.

ABSTRACT

Cryo-electron microscopy (cryo-EM) technique is widely used for protein structure determination. Current automatic cryo-EM protein complex modeling methods mostly rely on prior chain separation. However, chain separation without sequence guidance often suffers from errors caused by cross-chain interaction or noise densities, which would accumulate and mislead the subsequent steps. Here, we present EModelX, a fully automated cryo-EM protein complex structure modeling method, which achieves sequence-guiding modeling through cross-modal alignments between cryo-EM maps and protein sequences. EModelX first employs multi-task deep learning to predict Cα atoms, backbone atoms, and amino acid types from cryo-EM maps, which is subsequently used to sample Cα traces with amino acid profiles. The profiles are then aligned with protein sequences to obtain initial structural models, which yielded an average RMSD of 1.17 Å in our test set, approaching atomic-level precision in recovering PDB-deposited structures. After filling unmodeled gaps through sequence-guiding Cα threading, the final models achieved an average TM-score of 0.808, outperforming the state-of-the-art method. The further combination with AlphaFold can improve the average TM-score to 0.911. Analyzes conducted by comparing some EModelX-built models and PDB structures highlight its potential to improve PDB structures. EModelX is accessible at https://bio-web1.nscc-gz.cn/app/EModelX .

PMID:39394203 | DOI:10.1038/s41467-024-53116-5

J Chem Inf Model. 2024 Oct 11. doi: 10.1021/acs.jcim.4c01037. Online ahead of print.

ABSTRACT

The human immunodeficiency virus presents a significant global health challenge due to its rapid mutation and the development of resistance mechanisms against antiretroviral drugs. Recent studies demonstrate the impressive performance of machine learning (ML) and deep learning (DL) models in predicting the drug resistance profile of specific FDA-approved inhibitors. However, generalizing ML and DL models to learn not only from isolates but also from inhibitor representations remains challenging for HIV-1 infection. We propose a novel drug-isolate-fold change (DIF) model framework that aims to predict drug resistance score directly from the protein sequence and inhibitor representation. Various ML and DL models, inhibitor representations, and protein representations were analyzed through realistic validation mechanisms. To enhance the molecular learning capacity of DIF models, we employ a transfer learning approach by pretraining a graph neural network (GNN) model for activity prediction on a data set of 4855 HIV-1 protease inhibitors (PIs). By performing various realistic validation strategies on internal and external genotype-phenotype data sets, we statistically show that the learned representations of inhibitors improve the predictive ability of DIF-based ML and DL models. We achieved an accuracy of 0.802, AUROC of 0.874, and r of 0.727 for the unseen external PIs. By comparing the DIF-based models with a null model consisting of isolate-fold change (IF) architecture, it is observed that the DIF models significantly benefit from molecular representations. Combined results from various testing strategies and statistical tests confirm the effectiveness of DIF models in testing novel PIs for drug resistance in the presence of an isolate.

PMID:39393002 | DOI:10.1021/acs.jcim.4c01037

PLoS One. 2024 Oct 11;19(10):e0311897. doi: 10.1371/journal.pone.0311897. eCollection 2024.

ABSTRACT

With the increasing amount of data in railway communication system, the conventional wireless high-frequency communication technology cannot meet the requirements of modern communication and needs to be improved. In order to meet the requirements of high-speed signal processing, a high-speed communication signal processing method based on visible light is developed and studied. This method combines the adaptive equalization algorithm with deep learning and is applied to railway communication signal processing. In this research, the wavelength division multiplexing (WDM) and orthogonal frequency division multiplexing (OFDM) techniques are used, and fuzzy C equalization algorithm is used to softly divide the received signals, reduce signal distortion and interference suppression. The experimental results showed that increasing the step size could reduce the equalization effect, while increasing the modulation parameter will increase the bit error rate. Through deep learning to achieve channel equalization, visible light communication could effectively mitigate multi-path transmission and reflection interference, thereby reducing the bit error rate to the level of 0.0001. Under various signal-to-noise ratios, the system using the channel compensation method achieved the lowest bit error rate. This outcome was achieved by implementing hybrid modulation scheme, including Wavelength division multiplexing (WDM) and direct current-biased optical orthogonal frequency division multiplexing (DCO-OFDM) techniques. It has been proved that this method can effectively reduce the channel distortion when the receiver is moving. This study develops a dependable communication system, which enhances signal recovery, reduces interference, and improves the quality and transmission efficiency of railway communication. The system has practical application value in the field of railway communication signal processing.

PMID:39392828 | DOI:10.1371/journal.pone.0311897

Bioinformatics. 2024 Oct 11:btae574. doi: 10.1093/bioinformatics/btae574. Online ahead of print.

ABSTRACT

MOTIVATION: Predicting the properties of molecules is a fundamental problem in drug design and discovery, while how to learn effective feature representations lies at the core of modern deep learning based prediction methods. Recent progress shows expressive power of graph neural networks (GNNs) in capturing structural information for molecular graphs. However, we find that most molecular graphs exhibit low clustering along with dominating chains. Such topological characteristics can induce feature squashing during message passing and thus impair the expressivity of conventional GNNs.

RESULTS: Aiming at improving node features' expressiveness, we develop a novel chain-aware graph neural network model, wherein the chain structures are captured by learning the representation of the center node along the shortest paths starting from it, and the redundancy between layers are mitigated via initial residual difference connection (IRDC). Then the molecular graph is represented by attentive pooling of all node representations. Compared to standard graph convolution, our chain-aware learning scheme offers a more straightforward feature interaction between distant nodes, thus it is able to capture the information about long-range dependency. We provide extensive empirical analysis on real-world datasets to show the outperformance of the proposed method.

AVAILABILITY AND IMPLEMENTATION: The MolPath code is publicly available at https://github.com/Assassinswhh/Molpath.

SUPPLEMENTARY INFORMATION: Supplementary information are available at Bioinformatics online.

PMID:39392786 | DOI:10.1093/bioinformatics/btae574