Multivariate analysis and machine learning link sex and age with antibody responses to SARS-CoV-2 and vaccination

Intriguing relationships of antibody (Ab) response profiles and neutralization measures were found across a participant cohort followed since October 2020 and through subsequent vaccination periods. These findings were made using largely advanced machine learning, a form of artificial intelligence that allows computers to adapt and draw inferences from data without explicitly being programmed to do so. Most notably, the team found compelling evidence linking an individual’s biological sex and/or age with the ability to generate and maintain antibodies.

For this analysis, serum was collected from 970 participants and separated into groups by:

• Vaccination period (unvaccinated, dose 1, dose 2);
• Age (<40, 40-60, >60);
• Sex (male/female);
• SARS-CoV-2 infection acquired immunity status (positive or negative for IgG Abs against SARS-CoV-2 nucleoprotein);
• Study data and materials were managed through the CoVaRR-Net Biobank.

For each time point studied, nine different Ab responses were measured: IgG, IgA, and IgM titers specific for SARS-CoV-2 receptor binding domain (R), spike (S), and nucleoprotein (N), as well as neutralization measurements.

A multi-variate analysis was conducted on the acquired serology and neutralization assay data that could provide multiple possible groupings of the data, potential correlations between various data groups (both linear and non-linear) and applied various statistical functions to identify significant relationships.

Key findings:

• Age and sex are associated with overall antibody responses to vaccination and infection;
• Major differences in Ab responses between males and females are dependent on age and infection-acquired immunity;
• The correlation analysis showed quantitative links between sex and Ab responses;
• The correlation between Abs and neutralization measurements varies over time differently in males and females.

Overall, the researchers introduced an integrative and novel, multi-variate, machine learning, and network analysis-based tools, which can assess multiple immunological parameters over time and across a diverse population. Such an innovative analytical approach, supported by appropriate biobank infrastructure (CoVaRR-Net), is well suited to multiparameter data sets and longitudinal studies. It affords the ability to identify data behaviours that will inform complementary immune studies of this cohort, namely assigning the ever-important T-cell mediated response patterns, through which an integrated data analysis could finally identify the elusive knowledge of correlates of protection. The latter is critical to inform the design of future vaccines that could provide ultimate protection against infection, reducing morbidity and mortalities in very meaningful ways. Such an approach to immune response analysis across a cohort and over time can revolutionize how Ab and cell-mediated responses occur in response to emerging threats to human health, supporting a framework of cohort follow-up and study by an emerging national research biobank.

Read article