Replication Data for: On-Site Precise Screening of SARS-CoV-2 Systems Using an Attention-Based PLS-1D-CNN Model with Biomolecular Infrared Signatures

During the early stages of respiratory virus outbreaks, such as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the effective use of limited nasopharyngeal swabs for rapid and accurate screening is crucial for public health. In this study, we present a methodology that integrates attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR) with the adaptive iteratively reweighted penalized least squares (airPLS) preprocessing algorithm and a channel-wise attention-based partial least squares one-dimensional convolutional neural network (PLS-1D-CNN) model, enabling accurate screening of infected individuals within 10 minutes. Two cohorts of nasopharyngeal swab samples, comprising 126 and 112 samples from suspected SARS-CoV-2 Omicron variant cases, were collected at Beijing Youan Hospital for verification. To assess signal quality across different experimental procedures, we introduce a biomolecular importance (BMI) evaluation method, which quantitatively measures the significance of virus-related biomolecules in feature extraction, helping differentiate the quality of spectral signals collected under varying conditions. This approach reveals underlying biological correlations, facilitating the selection of higher-quality spectra and standardizing protocols to ensure consistent, high-quality spectral signal collection. For ATR-FTIR signals in cohort 2, which showed higher BMI, airPLS was used for signal preprocessing, followed by application of the channel-wise attention-based PLS-1D-CNN model for screening. Experimental results demonstrate that our model achieves a screening accuracy of 96.48%, sensitivity of 96.24%, specificity of 97.14%, F1-score of 96.12%, and an AUC of 0.99, meeting the World Health Organization’s recommended criteria for acceptable screening products.

在呼吸道病毒暴发早期，例如严重急性呼吸综合征冠状病毒2（severe acute respiratory syndrome coronavirus 2, SARS-CoV-2）暴发期间，高效利用有限的鼻咽拭子开展快速精准筛查，对公共卫生防控至关重要。本研究提出一种整合衰减全反射-傅里叶变换红外光谱（attenuated total reflection-Fourier transform infrared spectroscopy, ATR-FTIR）、自适应迭代重加权惩罚最小二乘（adaptive iteratively reweighted penalized least squares, airPLS）预处理算法，以及基于通道注意力的偏最小二乘一维卷积神经网络（channel-wise attention-based partial least squares one-dimensional convolutional neural network, PLS-1D-CNN）模型的研究方法，可在10分钟内完成感染者的精准筛查。研究团队在北京佑安医院采集了两队列鼻咽拭子样本，分别包含126份和112份疑似感染新冠奥密克戎变异株的病例样本，用于模型验证。为评估不同实验流程下的信号质量，本研究引入生物分子重要性（biomolecular importance, BMI）评估方法，该方法可量化提取特征中病毒相关生物分子的显著性，助力区分不同采集条件下光谱信号的质量。该方法能够揭示潜在的生物学关联，辅助筛选高质量光谱并标准化实验流程，以确保光谱信号采集的一致性与高质量。针对队列2的ATR-FTIR信号（该队列BMI值更高），研究采用airPLS进行信号预处理，随后应用基于通道注意力的PLS-1D-CNN模型完成筛查。实验结果表明，本模型的筛查准确率达96.48%、灵敏度96.24%、特异度97.14%、F1分数96.12%，曲线下面积（AUC）为0.99，符合世界卫生组织（World Health Organization, WHO）推荐的可接受筛查产品标准。