Supporting information for: New methods for analyzing serological data with applications to influenza surveillance

TL;DR

This paper introduces a biophysical model for influenza hemagglutination-inhibition assays and proposes new analytical methods, including SVD and multidimensional scaling, to interpret serological data more effectively.

Contribution

It presents a novel biophysical model of HI assays and new analytical techniques for interpreting serological titers, applicable to influenza surveillance.

Findings

Biophysical model clarifies assay mechanisms

New methods improve titer interpretation

Algorithm estimates missing titers

Abstract

For decades, the hemagglutination-inhibition (HI) assay has been used in epidemiological and basic biological studies of influenza viruses. The mechanistic basis of the assay results (called titers) is not well understood. The first part of this document describes a biophysical model of HI that illuminates the mechanistic basis of and provides the theoretical motivation for new ways of interpreting titers. The biophysical model is applicable to other serological assays; this fact is illustrated using the neutralization assay. The second part of the document describes some new ways of interpreting titers, which involve, among other methods, singular value decomposition and probabilistic multidimensional scaling. The third part of the document discusses biological and mathematical issues related to the determination of the effective dimensionality of titers, and describes an algorithm for recovering unavailable titers.