Using FIDA for in-solution binding in AI drug discovery: a conversation with Marissa Baker

Marissa Baker, whose background is in biochemistry and biophysics, leads the biochemical and biophysical screening at Genesis Therapeutics, an AI-driven drug-discovery company. In a conversation with Fidabio, Marissa Baker described where Flow Induced Dispersion Analysis (FIDA) fits in that work. This is a summary of what she said.

What technology to choose for difficult binding targets?

Genesis runs virtual screening through its AI platform, GEMS, which generates and ranks candidate compounds that chemists then triage, synthesise or purchase, and test in real assays whose data feed back into the model. Why did thy choose FIDA? It entered that loop as an in-solution way to generate dissociation constants and to confirm results across methods. The in-solution character is what drew Marissa Baker to it in the first place: existing in-solution options were either inconsistent, heavy on material, or dependent on a specialised technique that was hard to bring in house.

"The FIDA was exciting to have more of a quicker method that gave pretty reliable data. And was in solution," Marissa Baker said.

In solution technology for difficult binding targets

The capability that proved decisive was measuring binding in free solution rather than on a surface. Surface plasmon resonance (SPR) is the everyday workhorse at Genesis, but a few proteins never showed any ligand binding on it, even though binding was known from other methods, including structure. Run in solution on FIDA, those same proteins showed binding that matched the biochemical and cellular data, and the lesson, in Marissa Baker's account, was that measuring in solution mattered more than the team had realised.

That matters because Genesis deliberately takes on difficult-to-drug targets, whose protein reagents are often hard to make, with little precedent, and sometimes have to sit in a specific complex to behave. Here a second capability earns its place: because FIDA reports the size of the protein, its hydrodynamic radius, at the very concentration being tested, the team can check that a precious reagent is present in the form it should be, rather than an unexpected complex or aggregate, before reading anything into a binding result.

It also needs very little material per run. Because Genesis is a small-molecule company, proteins are labelled with a Cy5 dye to avoid compound interference in the UV; preparing that stock uses some protein, but each run then consumes very little and one stock lasts a long time, valuable for a hard-to-make reagent.

Where the work is heading

The plan is to move FIDA from confirmation toward front-line screening: higher-throughput setups that return a simple yes-or-no binding answer for brand-new targets from the AI platform, sampling many chemotypes quickly before confirming the hits with full affinity measurements.

What we learned from this conversation

• In free solution, no surface: revealed real, data-consistent binding on proteins where SPR showed nothing.
• Size at the tested concentration: the hydrodynamic radius reveals which molecular species is present, key for hard-to-make reagents and defined complexes.
• Sample-sparing: one Cy5-labelled stock supports many low-consumption runs.
• Fits an AI loop: generates dissociation constants that confirm and feed the GEMS workflow, and is scaling toward yes-or-no screening of new targets.

Curious how in-solution binding analysis could fit your discovery workflow? Explore the FIDA technology and the Fida instrument, or browse the peer-reviewed literature. Read more scientists’ experiences with FIDA.