A new preprint with @BrianCleary studying matrix completion – a technique to fill in missing measurements within a partially-observed matrix of values. If you have ever created or analyzed data with missing entries, this technique is for you. (1/5)
biorxiv.org/content/10.1101/…
One surprising result is that we can use measurements from individual antibodies to inform how the vast array of antibodies in our bodies behave (and vice versa). These two scales of the antibody response are notoriously hard to bridge. (4/5)
We also quantify the error of these predictions. In the influenza studies we examined, we predict missing values within measurement error. Had these studies utilized matrix completion, tens of thousands of experiments could have been saved. (5/5)
github.com/cleary-lab/Inhibi…