Lifting Biomolecular Data Acquisition

TL;DR

This paper introduces a neural compressed sensing approach that enables simultaneous measurement of multiple molecules, significantly increasing information density in wet lab experiments for ML-driven biological research.

Contribution

It proposes a novel neural extension of compressed sensing for function space, allowing concurrent molecule activity measurement and improved data efficiency.

Findings

Order-of-magnitude increase in information density.

Effective deconvolution of molecule-activity maps.

Validated on antibodies and cell therapies.

Abstract

One strategy to scale up ML-driven science is to increase wet lab experiments' information density. We present a method based on a neural extension of compressed sensing to function space. We measure the activity of multiple different molecules simultaneously, rather than individually. Then, we deconvolute the molecule-activity map during model training. Co-design of wet lab experiments and learning algorithms provably leads to orders-of-magnitude gains in information density. We demonstrate on antibodies and cell therapies.