TEPCAM: Prediction of T-cell receptor-epitope binding specificity via interpretable deep learning.

The recognition of T-cell receptor (TCR) on the surface of T cell to specific epitope presented by the major histocompatibility complex is the key to trigger the immune response. Identifying the binding rules of TCR-epitope pair is crucial for developing immunotherapies, including neoantigen vaccine and drugs. Accurate prediction of TCR-epitope binding specificity via deep learning remains challenging, especially in test cases which are unseen in the training set. Here, we propose TEPCAM (TCR-EPitope identification based on Cross-Attention and Multi-channel convolution), a deep learning model that incorporates self-attention, cross-attention mechanism, and multi-channel convolution to improve the generalizability and enhance the model interpretability. Experimental results demonstrate that our model outperformed several state-of-the-art models on two challenging tasks including a strictly split dataset and an external dataset. Furthermore, the model can learn some interaction patterns between TCR and epitope by extracting the interpretable matrix from cross-attention layer and mapping them to the three-dimensional structures. The source code and data are freely available at https://github.com/Chenjw99/TEPCAM.

A Multimodal Deep Learning Framework for Predicting PPI-Modulator Interactions.

Protein-protein interactions (PPIs) are essential for various biological processes and diseases. However, most existing computational methods for identifying PPI modulators require either target structure or reference modulators, which restricts their applicability to novel PPI targets. To address this challenge, we propose MultiPPIMI, a sequence-based deep learning framework that predicts the interaction between any given PPI target and modulator. MultiPPIMI integrates multimodal representations of PPI targets and modulators and uses a bilinear attention network to capture intermolecular interactions. Experimental results on our curated benchmark data set show that MultiPPIMI achieves an average AUROC of 0.837 in three cold-start scenarios and an AUROC of 0.994 in the random-split scenario. Furthermore, the case study shows that MultiPPIMI can assist molecular docking simulations in screening inhibitors of Keap1/Nrf2 PPI interactions. We believe that the proposed method provides a promising way to screen PPI-targeted modulators.

Machine learning of flow cytometry data reveals the delayed innate immune responses correlate with the severity of COVID-19.

Introduction: The COVID-19 pandemic has posed a major burden on healthcare and economic systems across the globe for over 3 years. Even though vaccines are available, the pathogenesis is still unclear. Multiple studies have indicated heterogeneity of immune responses to SARS-CoV-2, and potentially distinct patient immune types that might be related to disease features. However, those conclusions are mainly inferred by comparing the differences of pathological features between moderate and severe patients, some immunological features may be subjectively overlooked. Methods: In this study, the relevance scores(RS), reflecting which features play a more critical role in the decision-making process, between immunological features and the COVID-19 severity are objectively calculated through neural network, where the input features include the immune cell counts and the activation marker concentrations of particular cell, and these quantified characteristic data are robustly generated by processing flow cytometry data sets containing the peripheral blood information of COVID-19 patients through PhenoGraph algorithm. Results: Specifically, the RS between immune cell counts and COVID-19 severity with time indicated that the innate immune responses in severe patients are delayed at the early stage, and the continuous decrease of classical monocytes in peripherial blood is significantly associated with the severity of disease. The RS between activation marker concentrations and COVID-19 severity suggested that the down-regulation of IFN-γ in classical monocytes, Treg, CD8 T cells, and the not down-regulation of IL_17a in classical monocytes, Tregs are highly correlated with the occurrence of severe disease. Finally, a concise dynamic model of immune responses in COVID-19 patients was generalized. Discussion: These results suggest that the delayed innate immune responses in the early stage, and the abnormal expression of IL-17a and IFN-γ in classical monocytes, Tregs, and CD8 T cells are primarily responsible for the severity of COVID-19.

Effects of dialysis modality on mortality in patients with end-stage renal disease: A cohort study.

METHODS: Using a retrospective 15-year cohort, stratified by age, this study aimed to analyze the effect of dialysis modality on mortality of ESRD patients in a city of China. Study data were from the medical insurance information system of Kunshan, Jiangsu Province of China, and 1484 patients with ESRD, enrolled from 1 January 2005 to 31 December 2019 were included in this study. The primary outcome event was all-cause mortality, which was calculated in months. Dialysis modalities included hemodialysis (HD) and peritoneal dialysis (PD). Survival analysis and competing-risk regression model were performed in this study. RESULTS: HD costs significantly higher medical expense than the PD treatment regimen. The mean survival time was 121.28 (SE = 3.020) months for HD patients, while that was 94.68 (SE = 3.534) months for the PD. Ten-year survival rates of the young, middle-aged, and elderly were 0.82, 0.56, and 0.26, respectively. For the young (SHR = 0.869, 95% CI: 0.525-1.436) and middle-aged (SHR = 0.715, 95% CI: 0.484-1.057) ESRD patients, different dialysis modalities exhibited no statistical significance on the survival, but for the elderly, HD had a lower risk of mortality than PD (SHR = 0.747, 95% CI: 0.581-0.961). CONCLUSION: Survival of the young and middle-aged ESRD patients was superior to that of the elderly. Considering both survival time and direct medical costs, we recommend that PD could be a better choice for young and middle-aged ESRD patients, while HD may be suitable for older patients.

Economic evaluation of three dialysis methods in patients with end-stage renal disease in China.

OBJECTIVES: End-stage renal disease (ESRD) may result in different degrees of physical and psychological pain. Automated peritoneal dialysis (APD), continuous ambulatory peritoneal dialysis (CAPD), and hemodialysis (HD) as the main treatment methods lead to a heavy burden on social economic and family financial. However, there are few studies on the economic evaluation of the three dialysis methods in China. METHODS: Cost-effectiveness analyses were performed using Markov models based on longitudinal data for 15 years of different modalities in Kunshan City, China. Direct cost derived from medical insurance information system, and indirect cost referred to as loss of productivity. Sensitivity analyses were conducted to study uncertainty. RESULTS: The per capita total cost of CAPD was 664,027.00 yuan, the per capita utility is 5.9105. The per capita total cost of APD was 858,800.65 yuan, the per capita utility is 6.4548. The per capita total cost of HD was 1,281,213.64 yuan, the per capita utility is 6.1356. When CAPD was compared with HD, Incremental Cost-Effectiveness Ratio (ICER) was 1,323,389.53 yuan per QALY, compared with APD, ICER was 357,848.13 yuan per QALY. ICER value suggests that APD was cost-effective compared with CAPD and HD at a willingness-to-pay threshold of 538,200 yuan. CONCLUSION: Our research showed that APD is the most appropriate and HD is the worst in terms of cost-effectiveness. However, in fact, HD accounts for a high proportion in China, so some relevant policy suggestions need to be implemented to change the current situation.

Impact of dialysis modality conversion on the health-related quality of life of peritoneal dialysis patients: a retrospective cohort study in China.

BACKGROUND: To analyze the health-related quality of life associated with the conversion of dialysis modality among end-stage renal disease patients in China. METHODS: Patients were recruited from hospitals and a dialysis center in Kunshan, China. Patients converting from continuous ambulatory peritoneal dialysis to automated peritoneal dialysis were recruited as the observation group (n = 64), and patients continuing with continuous ambulatory peritoneal dialysis treatment were included in the control group (n = 64) after matching in this retrospective cohort study. Their health-related quality of life was measured using the kidney disease quality of life instrument in 2019 and 2020, respectively. Baseline socio-demographic characteristics and clinical data were collected in 2019. The before-and-after cross-group comparisons of subscale scores of two groups were conducted using a Student's t-test. Multiple linear regression models were fitted to identify the factors associated with the change of each scale. RESULTS: The health-related quality of life scores of the two groups was comparable in baseline, while the observation group had higher scores in Physical Component Summary (51.92 ± 7.50), Kidney Disease Component Summary (81.21 ± 8.41), Symptoms (90.76 ± 6.30), Effects (82.86 ± 11.42), and Burden (69.04 ± 15.69) subscales after one year. In multivariate regression analysis, the change of Physical Component Summary was significantly associated with conversion to APD (ß = 11.54, 95% CI [7.26-15.82]); the change of Mental Component Summary with higher education (ß =  - 5.96, 95% CI [-10.18--1.74]) and CCI (>2) (ß = 5.39, 95% CI [1.05-9.73]); the change of Kidney Disease Component Summary with conversion to APD (ß = 15.95, 95% CI [10.19-21.7]) and age (>60 years) (ß =  - 7.36, 95% CI [-14.11--0.61]); the change of Symptoms with CCI (>2) (ß = 7.96, 95% CI [1.49-14.44]); the change of Effects with conversion to APD (ß = 19.23, 95% CI [11.57-26.88]); and the change of Burden with conversion to APD (ß = 22.40, 95% CI [13.46-31.34]), age (>60 years) (ß =  - 12.12, 95% CI [-22.59--1.65]), and higher education (ß =  - 10.38, 95% CI [-19.79--0.98]). CONCLUSIONS: The conversion of dialysis modality had a significant impact on the scores of most subscales. Patients converting from continuous ambulatory peritoneal dialysis to automated peritoneal dialysis generally had improved health-related quality of life scores.

Do online applications for free assistive technology devices by individuals with disabilities introduce moral hazard? Evidence from Shanghai, China.

OBJECTIVES: To examine whether moral hazard may exist under unsupervised home-based online applications, leading to more assistive technology devices (ATDs) and larger per capita expenditures on ATDs than under supervised community center-based online applications. METHODS: Using the data from the Assistive Devices Resource Centre in Shanghai, descriptive statistics were estimated for the sociodemographics of applicants. Multiple linear regression and logistic regression were used to test the effect of the introduction of home-based online applications. RESULTS: In 2015-2016, there were marked increases of 22.3% in the total number of ATDs and 27.2% in the total expenditure on ATDs compared with 2013-2014. The per capita number and expenditure also demonstrated an increasing trend. More devices were applied for in 2015-2016 than in 2013-2014, yielding a higher expenditure per capita (P < .001). Interestingly, with an invisible price, more devices were applied for at home than in community centers (P < .001), but the expenditure per capita was smaller (P < .001). CONCLUSIONS: The introduction of online applications increased the number of ATDs per capita. The home-based applications induced the purchase of more ATDs but not higher expenditures on ATDs. Individuals with disabilities tend to request the maximum number of ATDs allowed by the application rules, which is an indicator of moral hazard. The prices of ATDs were not visible for individuals with disabilities, which may cause individuals to order costlier ATDs when applying at home. Stricter review may be needed to reign in the potential moral hazard among online applicants with disabilities.