Development and validation of multivariable prediction models of serological response to SARS-CoV-2 vaccination in kidney transplant recipients

Using machine learning for personalized prediction of longitudinal coronavirus disease 2019 vaccine responses in transplant recipients

The coronavirus disease 2019 pandemic has underscored the importance of vaccines, especially for immunocompromised populations like solid organ transplant recipients, who often have weaker immune responses. The purpose of this study was to compare deep learning architectures for predicting severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) vaccine responses 12 months postvaccination in this high-risk group. Using data from 303 solid organ transplant recipients from a Canadian multicenter cohort, models were developed to forecast anti-receptor-binding domain antibody levels. The study compared traditional machine learning models-logistic regression, epsilon-support vector regression, random forest regressor, and gradient boosting regressor-and deep learning architectures, including long short-term memory (LSTM), recurrent neural networks, and a novel model, routed LSTM. This new model combines capsule networks with LSTM to reduce the need for large data sets. Demographic, clinical, and transplant-specific data, along with longitudinal antibody measurements, were incorporated into the models. The routed LSTM performed best, achieving a mean square error of 0.02 ± 0.02 and a Pearson correlation coefficient of 0.79 ± 0.24, outperforming all other models. Key factors influencing vaccine response included age, immunosuppression, breakthrough infection, body mass index, sex, and transplant type. These findings suggest that artificial intelligence could be a valuable tool in tailoring vaccine strategies, improving health outcomes for vulnerable transplant recipients.

Serological responses and clinical outcomes following a three-dose primary COVID-19 vaccine schedule in kidney transplant recipients and people on dialysis

Objectives

Despite vaccination strategies, people with chronic kidney disease, particularly kidney transplant recipients (KTRs), remained at high risk of poor COVID-19 outcomes. We assessed serological responses to the three-dose COVID-19 vaccine schedule in KTRs and people on dialysis, as well as seroresponse predictors and the relationship between responses and breakthrough infection.

Methods

Plasma from 30 KTRs and 17 people receiving dialysis was tested for anti-Spike receptor binding domain (RBD) IgG and neutralising antibodies (NAb) to the ancestral and Omicron BA.2 variant after Doses 2 and 3 of vaccination.

Results

After three doses, KTRs achieved lower anti-Spike RBD IgG levels (P < 0.001) and NAb titres than people receiving dialysis (P = 0.002). Seropositive cross-reactive Omicron neutralisation levels were achieved in 11/27 (40.7%) KTRs and 11/14 (78.6%) dialysis recipients. ChAdOx1/viral-vector vaccine type, higher mycophenolate dose (> 1 g per day) and lower absolute B-cell counts predicted poor serological responses in KTRs. ChAdOx-1 vaccine type and higher monocyte counts were negative predictors in dialysis recipients. Among ancestral NAb seroresponders, higher NAb levels positively correlated with higher Omicron neutralisation (R = 0.9, P < 0.001). More KTRs contracted SARS-CoV-2 infection (14/30; 47%) than dialysis recipients (5/17; 29%) and had more severe disease. Those with breakthrough infections had significantly lower median interdose incremental change in anti-Spike RBD IgG and ancestral NAb titres.

Conclusion

Serological responses to COVID-19 vaccines in KTRs lag behind their dialysis counterparts. KTRs remained at high risk of breakthrough infection after their primary vaccination schedule underlining their need for booster doses, strict infection prevention measures and close surveillance.

Dynamic Profiling and Prediction of Antibody Response to SARS-CoV-2 Booster-Inactivated Vaccines by Microsample-Driven Biosensor and Machine Learning

Knowledge of the antibody response to the third dose of inactivated SARS-CoV-2 vaccines is crucial because it is the subject of one of the largest global vaccination programs. This study integrated microsampling with optical biosensors to profile neutralizing antibodies (NAbs) in fifteen vaccinated healthy donors, followed by the application of machine learning to predict antibody response at given timepoints. Over a nine-month duration, microsampling and venipuncture were conducted at seven individual timepoints. A refined iteration of a fiber optic biolayer interferometry (FO-BLI) biosensor was designed, enabling rapid multiplexed biosensing of the NAbs of both wild-type and Omicron SARS-CoV-2 variants in minutes. Findings revealed a strong correlation (Pearson r of 0.919, specificity of 100%) between wild-type variant NAb levels in microsamples and sera. Following the third dose, sera NAb levels of the wild-type variant increased 2.9-fold after seven days and 3.3-fold within a month, subsequently waning and becoming undetectable after three months. Considerable but incomplete evasion of the latest Omicron subvariants from booster vaccine-elicited NAbs was confirmed, although a higher number of binding antibodies (BAbs) was identified by another rapid FO-BLI biosensor in minutes. Significantly, FO-BLI highly correlated with a pseudovirus neutralization assay in identifying neutralizing capacities (Pearson r of 0.983). Additionally, machine learning demonstrated exceptional accuracy in predicting antibody levels, with an error level of <5% for both NAbs and BAbs across multiple timepoints. Microsample-driven biosensing enables individuals to access their results within hours of self-collection, while precise models could guide personalized vaccination strategies. The technology's innate adaptability means it has the potential for effective translation in disease prevention and vaccine development.

Letermovir Rescue Therapy in Kidney Transplant Recipients with Refractory/Resistant CMV Disease

(1) Background: CMV infections remain a problem after kidney transplantation, particularly if patients are refractory or resistant (r/r) to treatment with valganciclovir (VGCV) or ganciclovir (GCV). (2) Methods: In a single-center retrospective study, kidney transplant recipients (KTR) receiving letermovir (LTV) as rescue therapy for VGCV-/GCV-r/r CMV disease were analyzed regarding CMV history, immunosuppression, and outcomes. (3) Results: Of 201 KTR treated for CMV between 2017 and 2022, 8 patients received LTV following treatment failure with VGCV/GCV. All patients received CMV prophylaxis with VGCV according to the center's protocol, and 7/8 patients had a high-risk (D+/R-) CMV constellation. In seven of eight cases, rising CMV levels occurred during prophylaxis. In seven of eight patients, a mutation in UL97 associated with a decreased response to VGCV/GCV was detected. In four of eight patients, LTV resulted in CMV clearance after 24 ± 10 weeks (16-39 weeks), two of eight patients stabilized at viral loads <2000 cop/mL (6-20 weeks), and two of eight patients developed LTV resistance (range 8-10 weeks). (4) Conclusion: LTV, which is currently evaluated for CMV prophylaxis in kidney transplantation, also shows promising results for the treatment of patients with VGCV/GCV resistance despite the risk of developing LTV resistance. Additional studies are needed to further define its role in the treatment of patients with CMV resistance.

A machine learning model for predicting serum neutralizing activity against Omicron SARS-CoV-2 BA.2 and BA.4/5 sublineages in the general population

Supervised machine learning (ML) methods have been used to predict antibody responses elicited by COVID-19 vaccines in a variety of clinical settings. Here, we explored the reliability of a ML approach to predict the presence of detectable neutralizing antibody responses (NtAb) against Omicron BA.2 and BA.4/5 sublineages in the general population. Anti-SARS-CoV-2 receptor-binding domain (RBD) total antibodies were measured by the Elecsys® Anti-SARS-CoV-2 S assay (Roche Diagnostics) in all participants. NtAbs against Omicron BA.2 and BA4/5 were measured using a SARS-CoV-2 S pseudotyped neutralization assay in 100 randomly selected sera. A ML model was built using the variables of age, vaccination (number of doses) and SARS-CoV-2 infection status. The model was trained in a cohort (TC) comprising 931 participants and validated in an external cohort (VC) including 787 individuals. Receiver operating characteristics analysis indicated that an anti-SARS-CoV-2 RBD total antibody threshold of 2300 BAU/mL best discriminated between participants either exhibiting or not detectable Omicron BA.2 and Omicron BA.4/5-Spike targeted NtAb responses (87% and 84% precision, respectively). The ML model correctly classified 88% (793/901) of participants in the TC: 717/749 (95.7%) of those displaying ≥2300 BAU/mL and 76/152 (50%) of those exhibiting antibody levels <2300 BAU/mL. The model performed better in vaccinated participants, either with or without prior SARS-CoV-2 infection. The overall accuracy of the ML model in the VC was comparable. Our ML model, based upon a few easily collected parameters for predicting neutralizing activity against Omicron BA.2 and BA.4/5 (sub)variants circumvents the need to perform not only neutralization assays, but also anti-S serological tests, thus potentially saving costs in the setting of large seroprevalence studies.

When performance is not enough-A multidisciplinary view on clinical decision support

Scientific publications about the application of machine learning models in healthcare often focus on improving performance metrics. However, beyond often short-lived improvements, many additional aspects need to be taken into consideration to make sustainable progress. What does it take to implement a clinical decision support system, what makes it usable for the domain experts, and what brings it eventually into practical usage? So far, there has been little research to answer these questions. This work presents a multidisciplinary view of machine learning in medical decision support systems and covers information technology, medical, as well as ethical aspects. The target audience is computer scientists, who plan to do research in a clinical context. The paper starts from a relatively straightforward risk prediction system in the subspecialty nephrology that was evaluated on historic patient data both intrinsically and based on a reader study with medical doctors. Although the results were quite promising, the focus of this article is not on the model itself or potential performance improvements. Instead, we want to let other researchers participate in the lessons we have learned and the insights we have gained when implementing and evaluating our system in a clinical setting within a highly interdisciplinary pilot project in the cooperation of computer scientists, medical doctors, ethicists, and legal experts.