Accurate measurement of proteins and small molecules is fundamental to medical diagnostics and monitoring therapeutic response. Gold standard methods for quantitative measurements often rely on affinity reagents that exclusively bind to the target molecule to be measured. However, in practice, many affinity reagents also bind to off-target molecules, which is known as cross-reactivity. Cross-reactivity can lead to inaccurate quantification through false positives. This work develops a mathematical framework that corrects for cross-reactivity in measurements and provides a method to predict the precision of our measurement. This demonstrates the potential to expand the repertoire of affinity reagents useful for quantification and improve molecular measurement accuracy without having to change affinity reagents or design new assays.
This is the code-base accompanying the paper "Theoretical framework and experimental validation of multiplexed analyte quantification using cross-reactive affinity reagents", currently in pre-print on biorxiv.
data: Data used in the paper with metadata.explanations: Code used to generate explanatory plots/animations for the concepts of confidence intervals and cross-reactive binding curves.demos: Code that implements and demonstrates the methods from the paper without using real-world data.applications: Code that implements and demonstrates the methods from the paper with real-world data.cr_utils: Utility functions that implement the lower-level operations described in the paper.
For an affinity reagent that can bind to two target molecules, because of noise, many different combinations of target molecule concentrations can generate the same signal. In the following video, the red dot marks the true target molecule concentration mixture, and the region shaded in red marks all the target molecule concentrations that could have generated the same signal readout due to noise.