Drug discovery · Computational antibody design

Macromoltek

An Austin software company that has designed antibodies inside a computer since 2011, paid for by federal grants and customer fees rather than venture rounds. Its strongest public evidence is a decade old.

The intern who brought his own disease

In the fall of 2018, Monica Berrondo was in the San Francisco Bay Area raising money for her small Austin company when she was introduced to a junior high school student named Varun Thvar.

His mother had been struck by a sudden 104 degree fever. Diagnosis took a bronchoscopy: Mycobacterium abscessus, a fast growing bacterium that causes severe lung infections and shrugs off antibiotics. She responded at first, then the infection became resistant within four weeks. Nine months of intravenous antibiotics followed, and then a surgeon removed part of her lung.

The family had started a small foundation. Varun wanted to know why there was no antibody for this bug. Berrondo took the project, and took Varun on as a summer intern.

The work went like this, in her account. Model the target protein, run a deep learning model over its surface to map the places an antibody could grip, then search sequence and shape space on several scaffolds for a matching gripper. That produced tens of thousands of sequence and structure pairs, filtered by a second deep learning model. Two candidates came out of one summer. Neither has been tested in an animal in any public record.

The haystack problem

The standard way to get a therapeutic antibody is to make a mouse sick, wait for its immune system to answer, harvest what it produced, screen it, then rewrite the winner to look human enough to inject. That last step can break the thing that made the antibody work. The cycle is measured in years.

Macromoltek sells the alternative: design the human antibody from the target structure outward, and make only the handful worth making. The company has been clear about the limit: its designs are only as good as the algorithm producing them, and everything still has to be tested at a bench.

The field is now crowded with better funded rivals. Absci said in January 2023 that it had created and validated de novo antibodies with zero shot generative AI, backed by a wet lab it says tests nearly three million designs a week. Generate Biomedicines filed for a Nasdaq listing in February 2026 with an antibody in Phase 3, 312 employees and a $223.0 million net loss for 2025. Macromoltek lists two people on its leadership page.

A mother, a daughter, and a Rosetta apprenticeship

Berrondo did her doctorate in chemical engineering at Johns Hopkins on an NIH Ruth L. Kirschstein fellowship, studying how mutations change protein structure and function. Her publications from those years are a tour of the Rosetta software world: domain insertion prediction in 2008, a docking benchmark and the ROSETTA3 suite paper in 2011, and a 2012 paper on supercharged, heat resistant antibodies led from the Center for Systems and Synthetic Biology at the University of Texas at Austin.

Her mother, Susana Kaufmann, was retiring from a career writing software for financial services. Kaufmann had taught her daughter to program; twenty years later the daughter taught her mother biology, and Kaufmann became the company CTO. A third family name, Manuel Berrondo, appears on the 2014 modeling paper and on the patent application as an inventor.

From predicting to designing

The first product was a website: upload a sequence, get a structure back in hours instead of the weeks a crystallographer needs. In 2014 that engine, SmrtMolAntibody, was entered in the second Antibody Modeling Assessment, a blind test in which groups receive eleven antibody sequences and submit predictions before the crystal structures are revealed. Ten of the eleven models came back in good agreement, averaging 1.4 angstroms.

The National Science Foundation paid for the next decade of engineering: $178,700 in 2015 to automate humanization, then $1,193,494 from 2016 to 2020 to automate design. By the Winter 2018 Y Combinator batch the pitch had shifted from software subscriptions to service work, and TechCrunch reported the demo day numbers: ten paying customers, $50,000 of revenue for 2017, and $500,000 in letters of intent.

Around 2017, by Berrondo’s account on a 2020 podcast, the company crossed from improving existing antibodies to designing new ones. Its last peer reviewed paper, in April 2020, is a survey of deep learning methods for antibodies.

Then the pandemic, then quiet

On April 1, 2020, Macromoltek announced de novo design aimed at COVID-19. Six days later Cytovia Therapeutics named it as designer of the SARS-CoV-2 arm of a bi-functional natural killer cell engager, with Cytovia’s CEO saying the goal was a clinical trial by the end of 2020. No clinical record for that program has appeared. The NSF Phase II ended in August 2020. The Austin Community College Bioscience Incubator, where the bench work was done, now lists the company among alumni. The 2018 website listed a dozen staff; the current one lists the two founders.

What is proven, and what is still claimed

Read plainly: the modeling engine was tested once against withheld crystal structures, in 2014, and did well. What came after, the design platform and the machine learning layer, has been described by the company but never put through a comparable public test. The customers were real enough to produce revenue in 2017, and not one has been named since.

What to watch

• A named partner, or a designed molecule appearing in someone else’s pipeline. That is the piece the record is missing.
• Any peer reviewed result after 2020, or an entry in a blind benchmark like the 2014 one.
• A new federal award. Nothing has landed since the Phase II ended in August 2020.
• Any securities filing at all. There has never been one.

In their words

Related companies

• Abalone Bio: Shared investor: Company-Stated Partners. An East Bay biotech that screens antibodies for what they do to a receptor, not how tightly they stick to it
• Nabla Bio: Also in Drug Discovery and Delivery. A Cambridge company that writes antibodies on a computer and tests them in its own lab, aimed at the cell surface proteins conventional discovery keeps missing
• Diffuse Bio: Also in Drug Discovery and Delivery. A Y Combinator company founded by one of the researchers who first pointed diffusion models at proteins
• Adaptyv Bio: Also in Drug Discovery and Delivery. A Swiss automated laboratory that turns AI designed protein sequences into measured binding data, and that has published the full results of three open design

Sources

1. Public recordNSF SBIR awards 1520463 and 1632399NSF and USAspending · 2015 to 2020
2. Public recordAutomated Aufbau of antibody structures using SmrtMolAntibodyProteins · Aug 2014
3. Public recordSecond antibody modeling assessment (AMA-II)Proteins · 2014
4. Public recordA Review of Deep Learning Methods for AntibodiesAntibodies (Basel) · Apr 28, 2020
5. Public recordUS patent application 2014/0100834 A1, applicant MacromoltekUSPTO via Justia · Apr 10, 2014
6. Public recordSEC EDGAR, NIH RePORTER, USAspending and Europe PMC searchesFederal databases · Sep 23, 2026
7. IndependentMother-daughter duo joins Y Combinator: MacromoltekVentureBeat · Mar 8, 2018
8. Independent64 startups at the Y Combinator W18 Demo DayTechCrunch · Mar 19, 2018
9. IndependentAustin startup Macromoltek and its antibody modeling softwareAustin American-Statesman · Apr 20, 2013
10. InterviewInterview: Rising Star Monica Berrondo from MacromoltekUnder30CEO · May 8, 2013
11. InterviewAntibodies with Dr. Monica Berrondo, Science in the Mall Y’allfounding_media · 2020
12. IndependentCurrent and alumni member companiesACC Bioscience Incubator · Sep 2026
13. CompetitorAbsci validates de novo antibodies with zero-shot generative AIGlobeNewswire · Jan 10, 2023
14. IndependentGenerate Biomedicines files for Nasdaq IPO, S-1 summaryTradingView · Feb 4, 2026
15. CompanyPatient advocacy leads to synthetic antibodies for mycobacteriaGEN Biotechnology · Feb 2023
16. CompanyMacromoltek de novo antibody design offering targeting COVID-19PR Newswire · Apr 1, 2020
17. PartnerCytovia and Macromoltek to develop dual-acting NK immunotherapyGlobeNewswire · Apr 7, 2020
18. CompanyMacromoltek website, team page and 2018 about page archivemacromoltek.com and Internet Archive · Sep 2026