Jul 30, 2026 5:57 AM
From Compound Microarray to a Global Drug Discovery CRO: A Conversation with Haiching Ma of Reaction Biology
Participants: Haiching Ma, PhD (Founder & CSO, Reaction Biology), Jason Amsbaugh, MBA (CEO, Samba Scientific), Trent Carrier, PhD, MBA (CEO, Carolina Molecular)
Overview
Reaction Biology is celebrating its 25th anniversary this year, having grown from a small technology company experimenting with small molecule compound microarrays at nanoliter scale into a global CRO with four campuses across the US and EU, roughly 200 employees, and a catalog of more than 2,000 biological targets, covering broad drug discovery modalities, including small molecules, protein degraders, ADCs, and biologicals. We sat down with founder and CSO Haiching Ma to talk about how the company got its start, how it has scaled through acquisition, where AI genuinely is (and isn’t) changing drug discovery today, and what he thinks Reaction Biology needs to become over the next 25 years.
A Technology Company First
Jason Amsbaugh: I’d like to start with the founder story for Reaction Biology. You now offer everything from biochemical assays to in vivo studies. Where did you start, and how did you get going?
Haiching Ma: We started the company in 2001, so this year is our 25th anniversary. When we started, we weren’t really a CRO. We were more of a technology company. This was around the time DNA and protein microarrays were new, and everyone was working on that technology. We had the idea to apply it to small-molecule compound microarrays instead, because high-throughput screening was extremely popular at the time: every major pharma had its own HTS facility, and some cost a couple hundred million to build. I visited a few of them and thought, we could do this kind of screening in a box, using our small-molecule microarray technology licensed from UPenn, instead of building an entire facility.
We pitched many large pharma companies on spotting their compound libraries into a DNA-microarray format and running reactions at nanoliter scale instead of microliter scale. It was a great idea on paper, but nanoliter chemistry behaves very differently than larger-volume reactions: it took enormous effort just to keep small molecules from crashing out of solution. The technology was probably a little ahead of its time, and we couldn’t generate enough customers on that platform.
So, we pivoted: we took what we’d learned from the nanoliter droplet and microfluidic work and applied it to a more moderate scale, larger than nanoliter droplets but still far more efficient than the industry standard, to save customers on reagents and money. That’s how we built our first kinase profiling service, HotSpot™. At the time, the gold standard was a radioisotope-based filter-binding assay, accurate but wasteful; one customer told us their old process generated gallons of radioactive wastewater. We moved that same radiometric chemistry into our miniaturized system, using only a small fraction of the radioisotope. That became the foundation of Reaction Biology’s kinase business, which today serves hundreds of customers a year. Through all these years, we have played important roles helping many of our clients to move their kinase drugs into clinical trials and even gain FDA approval.
From there, we kept expanding into new target classes. Around 2008, we were the first company to build epigenetics services, at a time when almost no one was offering that commercially. We’ve kept adding capabilities ever since, and today, Reaction Biology has become a go-to CRO for over a thousand companies and academic labs every year when they consider new drug discovery programs, due to our large drug target collections and broad capabilities.
Scaling Through Strategic Acquisition
Jason Amsbaugh: Is the majority of your work still kinase profiling, or has it branched out?
Haiching Ma: We’ve expanded from biochemical assays into cell-based and biophysical assays, in vivo studies, and toxicology, including a GLP/GMP service line. We now operate four campuses.
In 2019 we acquired ProQinase in Germany, which brought in vivo, immunology, and additional cell biology capabilities. In 2022 we were acquired by a private equity firm, and we’ve continued to grow since: adding a company in Hershey, Pennsylvania focused on in vivo cancer models, and a GLP-focused company in Germany that works on larger-molecule therapeutics like antibodies and hormone therapies, closer to batch-release testing than early discovery, though we can serve a much broader spectrum of customers. The vision behind all of this has been to grow Reaction Biology into a globalized CRO that covers everything from early discovery through clinical trial support. We’re still building toward that. Last year we added a full set of in vitro safety panels to help our customers predict their drug candidates’ adverse events and lower clinical attrition.
Jason Amsbaugh: I knew you had four global sites, but I didn’t realize that growth happened mostly in the last five to ten years, through acquisition rather than organic growth.
Haiching Ma: Right. Both organic growth and M&A are strategies to grow the company. Organic growth is healthier in some ways, for example, we added biophysical and cell biology mainly through organic growth, but it’s slow. With financial backing, M&A becomes an option depending on your strategy. We first acquired ProQinase in 2019 specifically to get a foothold in Europe, adding functional assays and additional immuno-oncology capabilities. Not long after, the pandemic started and pharma R&D shut down internally; we saw a period of fast growth as more work moved to CROs. Right after the pandemic, the market for M&A was very active, and we were acquired ourselves.
Building One Global Culture
Jason Amsbaugh: Across all of these acquisitions, what do you think matters most for a successful cultural integration, honoring each site’s own culture while building one Reaction Biology identity?
Haiching Ma: As a CRO, success starts with the science: the quality of service for your customers. Even as ownership changes, the service line and quality can’t slip. We still need to always add consultative value to our customer relationships. Culturally, there are real differences between our European and US teams (different vacation norms, different working styles), and you must recognize and respect those differences rather than try to erase them. At the same time, you must keep reinforcing that it’s one Reaction Biology company: it’s not one site’s revenue, it’s the whole company’s revenue, and success is shared. Regular touchpoints (town halls, people introducing themselves and their backgrounds) help a lot. At around 200 employees, we’re still small enough that people can get to know each other, and the more they understand each other’s backgrounds, the better they work together.
Competing in a Global CRO Landscape
Jason Amsbaugh: With the increased scrutiny on Chinese CROs around national security and genomic data, has that shifted your business or created new opportunities?
HaichingMa: Youdefinitelyseegeopoliticsaffectingscience. ReactionBiologyhasa large academic and NIH customer base: we’ve been part of the NCI Chemical Biology Consortium since 2016, and I’ve watched the membership shift firsthand: a few years ago, that consortium included CROs providing services out of China; today, federally funded work has to be done in the US or Europe if no one else provides the services needed.
That said, it is public knowledge that the majority of major Chinese CROs’ business is still coming from US customers who aren’t using government funding and are free to work with whoever offers the best economics, and the cost difference is real. So, there’s genuine competition, but everyone has their own strengths. Reaction Biology’s edge is always staying ahead of the discovery game and the breadth of science it has built through these years: we’ve built a library of more than 2,000 biological targets, plus a wide range of cell-based assays and unique in vivo models, which is hard for any single competitor to match. There are also areas where regulation works in our favor: both HHS and FDA are actively trying to bring clinical trials back to the US. HHS has recently launched “Operation TrialBlazer” to restore America’s leadership in clinical research. That’s actually an area where we’re still building our own capabilities, and a focus for us going forward.
Cutting Through the AI Hype
Trent Carrier: What do you see with the emergence of AI models in drug discovery? Is that going to change the economics of this market?
Haiching Ma: It’s coming, for sure, but right now, I’d say AI drug discovery is still more theory than proven practice. A lot of companies use “AI” as a buzzword to raise funding, and the failure rate we see is quite high: we’ve worked with well-funded AI companies whose compounds test no better than a traditional medicinal chemistry approach. That said, there’s so much investment going into this space that success is coming.
Traditionally, if a customer is developing a kinase inhibitor, they might synthesize 300 to 500 analog compounds to find the best candidate. AI-guided design can shrink that number substantially, so in that narrow sense, it could shrink the amount of wet-lab testing we do per program.
Trent Carrier: Does AI end up expanding the number of molecules you’re seeing, because it generates more chemistry in silico, or does it reduce that number because it pre-screens more effectively?
Haiching Ma: Honestly, both. For a specific, well-characterized target, in silico screening shrinks the number of compounds that need wet-lab testing. But it’s also opening up entirely new target areas that were previously out of reach, because traditionally you needed a crystal structure and deep prior knowledge of a target to work on it at all. Now AI-driven groups are willing to take on targets nobody had tools for before, we have one customer sending us 20 to 30 new targets a month. That’s a real opportunity for us to build new capabilities, even though developing assays for genuinely novel targets isn’t easy.
Jason Amsbaugh: But all of that still has to be validated in a wet lab: there’s likely still a lot of variability across cell lines, protein mutants, and assay conditions that makes any AI-driven prediction highly dependent on how the underlying assay is designed.
Haiching Ma: Exactly right. A crystal structure, or an AI-predicted structure, is just one static snapshot of a protein, but proteins are dynamic, and small changes in the cellular environment change how they fold and function. That complexity is a big part of why pure AI prediction hasn’t fully succeeded yet: it captures one side of the picture and misses the other. There’s still a long way to go before AI fully replaces the wet lab.
The Shift Toward Multi-Omics and Global Biology
Trent Carrier: The other big trend we’re seeing is multi-omics. How is that showing up in the work you’re doing?
Haiching Ma: For twenty years, precision medicine has focused on one target at a time (a single kinase, for instance), which is why kinase inhibitors have boomed. But a kinase isn’t active in just one pathway; it touches many proteins across many pathways, so understanding the global effect matters just as much as understanding any one target. One of the best examples could be the Science paper (DOI: 10.1126/science.ads7152) published by St. Jude scientists last year on the phosphorylation pattern of RNA polymerase II. Using our HotSpot™ platform, we have identified over 100 kinases that can phosphorylate Pol II at different stages, contrary to the traditional view that only a handful of CDKs phosphorylate Pol II.
That’s genuinely hard: block one pathway, and another may compensate and keep a cancer growing, just more slowly. There’s also a bit of a built-in tension here, because as inhibitors get more selective to reduce side effects, that doesn’t necessarily mean broader clinical applicability. A drug that hits multiple relevant pathways may actually have a better therapeutic window. Multi-omics approaches are therefore important, helping build a more complete view of how the drugs work in cell or animal models.
Trent Carrier: Do newer model systems, like spheroids and organoids, start to address that more global, systems-level question?
Haiching Ma: That’s the hope, but so far, no organoid platform has been broadly adopted as a universal model: you might have a good model for lung cancer, or breast cancer, but nothing that fits everything yet. Given regulatory pressure to reduce animal testing, there’s a lot of government-funded effort going into this, even 3Dprinted organ and organoid models. It’s promising, but still a long road.
Trent Carrier: It’s interesting: the field went from broad to very precise, down to a single kinase binding pocket, and now it seems to be expanding back toward global, systems-level thinking. Is that a fair read?
HaichingMa: I think that’s just the natural progression of science. Early on, without genomic tools, people looked at broad cellular effects because that’s all they could measure. As genomics and crystal structures matured, the field narrowed in on very specific targets, because that’s what the tools allowed. Now, with multi-omics tools and whole-tissue models becoming possible, the field is circling back to the bigger picture again. I think you ultimately need both.
Precision Chemistry: Molecular Glues, PROTACs, and Resistance
Jason Amsbaugh: On the design side, are people moving toward multi-kinase inhibitors that intentionally hit redundant pathways, rather than pure single-target specificity?
Haiching Ma: Right now, most programs still aim for a selective inhibitor per target, then combine multiple compounds for combination therapy. Designing a single molecule to deliberately hit two or three specific kinases across different pathways, and nothing else, is a much harder problem than it sounds: early kinase inhibitors were often broadly active almost by accident, because researchers didn’t yet have the structural knowledge to make them selective. Doing the reverse (precisely engineering multi-target selectivity into one molecule) hasn’t really been solved yet.
Jason Amsbaugh: Given how much acquired resistance shows up in cancer, are people using wet-lab or AI approaches to anticipate resistance mutations ahead of time, rather than reacting after the fact?
Haiching Ma: As far as I know, there’s no real technology that predicts which resistance mutation will emerge: it’s still mostly reactive. When a patient develops resistance, you sequence to identify the responsible mutation or fusion protein, then work backward to find or develop something that addresses it. We published a Nature Biotech paper in April (DOI: 10.1038/s41587-026-03090-8), in collaboration with Fred Hutch, profiling roughly 100 FDA-approved kinase drugs against about 800 kinases, including known resistance mutations: the idea being that some already-approved drugs may happen to work against a specific resistance mutation, even if that wasn’t their original intent, which matters a lot for rare mutations where developing a new drug isn’t commercially viable. We’re also working informally with a few researchers on a newly identified kinase target tied to cancer metabolism that has no approved therapy yet.
Jason Amsbaugh: That repurposing approach seems especially valuable in something like pediatric cancer, where treatment options are limited.
Haiching Ma: Yes, the goal is to give clinicians a data-backed starting point for off-label use when nothing else is approved for a given rare cancer.
Trent Carrier: Do you see molecular glues as a true platform approach, or more of a case-by-case discovery process?
Haiching Ma: I actually think molecular glues are a true platform that may benefit more from AI than PROTACs do. A PROTAC is fundamentally a linker chemistry problem (two molecules joined together), which is a space people already understand reasonably well. A molecular glue must accomplish two functions within one much smaller molecule, which is a genuinely harder rational-design problem, and one where AI-assisted design could help. Once you do have a working molecular glue, though, it behaves like a normal small molecule for delivery purposes, which is a real advantage over the larger PROTAC format, and that’s part of why we’re seeing more interest in glues.
Trent Carrier: What about DDR and other kinase-adjacent pathways? Any resurgence there on the small-molecule side?
Haiching Ma: Yes, we do see there is more research in the past 2 or 3 years focusing on drug development around the DDR and kinase-adjacent space, especially for targets such as Wee1 and PKMYT1. We’re also seeing more combination-therapy approaches rather than single molecules engineered to hit multiple targets in these pathways; combining two selective compounds is still a very practical path. Transcription factors related to the DDR and other kinase-adjacent pathways are also becoming an interest. The idea of blocking transcription itself, rather than one upstream protein, is appealing, but transcription factors remain a genuinely difficult target class. We’re still building dedicated assays in that space, with some early success, but no general platform yet.
What’s on the Roadmap
Trent Carrier: With a library of more than 2,000 targets, what’s on your roadmap for the next six to twelve months?
Haiching Ma: Since we’re a service provider rather than a drug discovery company ourselves, we build what the market is asking for. We’re seeing a growing share of requests unrelated to cancer (metabolic disease, neuroscience, and weightmanagement targets), so we’re deliberately diversifying our therapeutic-area coverage beyond oncology, even though cancer remains our core business and the largest area in the industry overall.
Trent Carrier: What core competencies from your oncology work let you move into these new areas quickly?
Haiching Ma: Our real advantage is scientific know-how in assay development, biochemistry, and biophysics: that expertise is largely target-agnostic, even if we’ve focused it on oncology because that’s where demand has been. On the oligonucleotide side (antisense, siRNA), I don’t think the barrier is biology anymore, it’s delivery, which is more of an engineering problem than a science problem, like what we saw solved for mRNA vaccines. And despite how much attention ADCs get, I don’t think small molecules are going anywhere: patients still generally prefer a pill over an infusion, and even ADCs rely on small-molecule payloads that need to be synthesized.
Other than continuing to add new capabilities and targets internally, we also need to collaborate with other leading CROs to provide customers with full services that will help them improve their success rate. For example, there is an education gap around sequencing and omics data earlier in development. In principle, more sequencing during cell-based and in vivo work would give better toxicity predictions, but in practice it’s often skipped, partly because sequencing has historically been far more expensive than a standard biochemical assay. As sequencing costs continue to fall, I expect it to become just another standard tool alongside the assays we’ve always run. That is how Carolina Molecular can help.
Looking Ahead to the Next 25 Years
Trent Carrier: Twenty-five years in, what does the next 25 look like for Reaction Biology?
Haiching Ma: If Reaction Biology is still Reaction Biology in another 25 years, I think we’ll need real AI drug discovery capability of our own, whether we build it or adopt it through partnerships: not every company will build great AI in-house, so AI is increasingly going to be delivered as a service, woven into every part of what we offer, potentially helping customers identify targets and advance compounds, not just test them.
We’ll also need new model systems as regulators push to reduce animal testing (2D and 3D cell models, organoids, organ-on-a-chip), built up step by step rather than all at once, and we’re in active conversations about which of these technologies to adopt. We’re also still building capabilities like DMPK, toxicology, and clinical sample support here in the US, since increasingly, for regulatory and geopolitical reasons, more work needs to happen domestically rather than internationally. There’s a lot to do.
Conclusion
Twenty-five years after starting as a small technology company betting on small molecule microarray, Reaction Biology has grown into a global CRO spanning early discovery through toxicology and clinical support, while staying anchored to the same core idea: deep, target-agnostic scientific expertise applied wherever the science, and the market, is heading next. As Haiching’s answers make clear, that means treating AI as a tool to build with rather than a threat to build against, investing in the model systems regulators are asking for, and continuing to expand into the therapeutic areas and geographies its customers need most.
At Carolina Molecular, we’re proud to work alongside partners like Reaction Biology who share our focus on getting the right treatment to the right patient at the right time. We invite you to follow along as Reaction Biology continues building toward its next 25 years.