How to fit tens of thousands of virus neutralization curves

Jesse Bloom

Fred Hutch Cancer Center / HHMI

This talk is about a methodology from the CEIRR computational core

This talk does not focus on a biological question.

It describes the computational methodology backing a new approach; the approach and its application to biological questions will be presented by Caroline Kikawa at the trainee spotlight on Wednesday.

Neutralization curves are fundamental to measuring antibody immunity

We recently developed way to measure these curves in much higher throughput

A single 96-well plate can measure ~3,000 curves (~1,000 titers in triplicate)

Here I will just describe the analysis

Component 1: Python package that fits neutralization curves

Component 2: pipeline that processes data to fit tens of thousands of curves

https://github.com/jbloomlab/seqneut-pipeline

Input 1: list of viral barcodes

Input 2: sample and sequencing data for each well of 96-well plate

Input 3: overall configuration file

https://github.com/jbloomlab/flu_seqneut_H3N2_2023-2024/blob/main/config.yml

Output: all curves for each sera

Output: numerical titers for all sera

Summary

We now have a method by which one graduate student can measure ~10,000 neutralization titers in a few months

However, analyzing the data is harder than just looking at a hemagglutination-inhibition assay plate

We hope the computational methods described here make the analysis easy enough that many labs will start using this approach