The human microbiome – the trillions of bacteria, archaea, fungi, and viruses that live in and on us – acts like a distributed organ, shaping metabolism, immunity, and even how we respond to therapies. As sequencing has become routine and analysis more sophisticated, shotgun metagenomics and multi-omics readouts are moving the field from correlations to actionable insights, with AI increasingly used for pattern discovery and decision support. The conversation below explores where this is already paying off and what’s likely next.

Manoj Dadlani is Executive Chairman and Chief Commercial Officer at Cmbio, the integrated microbiome services platform formed from Clinical Microbiomics, CosmosID, MS-Omics, and DNASense. He previously served as CEO of CosmosID. Manoj’s background spans management consulting and company building, and he holds undergraduate and graduate degrees in biological engineering from Cornell University.

‍

Manoj, thanks for joining us. Since we last spoke in 2019, a lot has changed. What advancements have you seen over these past five years in the field at large and in any specific application areas of interest to you?

A lot has changed since 2019. I think of it in a few categories.

Starting with pharma, a couple of drugs have gone through FDA clearance – two for C. difficile: one from Seres Therapeutics and one from Rebiotix (now Ferring Pharmaceuticals). Those were the first microbiome therapeutics to get approved. We also saw positive Phase 3 data from MaaT Pharma for a graft‑versus‑host disease therapeutic targeting the gut. That’s one of the most exciting to me because it’s outside C. diff – that’s a great place to be, as moving into systemic applications opens the door to other modalities and disease states. Vedanta is going into Phase 3 as well (also C. diff). There are other drugs and programs in earlier stages across IBD and additional areas. The more of those that progress, the better for the field.

Now looking at it from the consumer side, thinking about probiotics and testing, there’s much more emphasis on running clinical trials in the consumer space. Consumers are starting to value science‑backed products – still early days, but there’s been progress. On the testing side, since the uBiome disaster, there’s been a shift in direct‑to‑consumer testing toward shotgun metagenomics. We’re working with many companies there to improve data quality. It’s still the early days in terms of actionability and clinical validation of recommendations, but the push toward better methods is encouraging, and I hope the regulatory landscape follows so that we see more clinical use.

In academic research, there’s a lot of exciting work, particularly in oncology and on drug response. Predicting whether a drug will work based on the microbiome is, to me, one of the most exciting areas to watch. It’s practical, and you could design a good regulatory framework to validate something like that, including getting it through the FDA. It’s definitely an area to watch – there’s so much research coming out. Trials are getting bigger, and the data is still relatively shallow, comparing the depth in cancer genomics to microbiome datasets, but as that grows, we can do more. AI is accelerating learning every month, and helping us find associations to build better diagnostics and therapeutics. It’s changing constantly, and it’s super exciting.

Last time you spoke with us, you mentioned that some of the best applications would use machine learning and AI to understand microbiome profiles. How has your perspective changed over the last five years? How are bioinformatics and AI already impacting the field, and how will they continue to?

In the last three of those five years we’ve been building infrastructure. Data quality is critical for these models. There’s been a big shift toward shotgun sequencing with better datasets, driven by lower sequencing costs – it’s hard to justify not doing shotgun sequencing. There’s still a place for 16S depending on the study, but this infrastructure shift has paved the way for the next wave of AI/ML and deep learning.

On bioinformatics tools in general, databases have improved – their quality and naming have gotten better, which enables better tools. Many tools are now in third, fourth, fifth iterations; accuracy keeps improving. We’re also seeing more commonality in which tools people use, which helps compare data more easily.

On AI, the landscape changes monthly. It’s insane. The next wave we’re hearing about and starting to work with is digital twins, which are in‑silico simulations of interventions or disease states to understand how to improve outcomes, and to learn when we make things worse by accident. For example, the University of Chicago developed an infant digital‑twin model called Q-net trained on thousands of infants sampled weekly for the microbiome, aiming to see whether early microbiome trajectories can predict neurological disease risk in children around two years of age. More broadly, AI is shifting from pattern recognition to decision support. That’s where the real power may come – acting as an additional “brain” to help us make better decisions.

What are you most excited about in bioinformatics tools in the coming years, and what remains to be done?

I was just having a conversation about this yesterday. I’d say it’s Personalization. It’s one thing to predict dysbiosis or elevated inflammation and a higher risk of, say, kidney disease; the real question is, what do you do with that prediction? I’m excited about tools that translate insights into tailored interventions – nutrition programs first, before drugs – to prevent these diseases from coming onboard. To me that’s the most exciting, because that’s the best for humanity.

The other side is medication response: how to prepare the body to maximize efficacy and minimize adverse effects. As the tools mature, I think that’s going to change the game.

You had previously said that “pharma needs some wins,” and we’ve seen highs – new FDA approvals, including the first oral fecal microbiota product – and lows – such as investor caution (e.g. related to 4D Pharma going under) and a lack of approvals outside C. diff and other GI indications. How have these headwinds impacted your expectations, and how optimistic are you about the microbiome approach going forward?

It depends on the lens you’re looking at. Short‑term, it’s painful – especially for us as a service provider. If companies can’t raise or go under, that impacts our business. Stepping back, in about 10 years we’ve taken two drugs through the FDA in an entirely new modality. That’s exciting. We had to create a regulatory framework, work within it, and get drugs through it. Now there’s a pipeline targeting tougher conditions, so I’m still very optimistic. We need to see programs reach the finish line, but a lot has happened in the past decade.

What lessons have you learned from those headwinds and from watching the space evolve so quickly?

For trials that haven’t panned out, product design may have needed more insight; some teams may have jumped the gun. Also, don’t over‑interpret shallow data: early findings from 16S don’t always hold at higher resolution. Invest in high‑quality, deeper data.

There are management lessons too in terms of running lean and efficient to hit milestones. Drug development is long and hard; it’s high risk but a big win if you get there. Having the right team and advisors, using the best data, and resisting the urge to over‑interpret are critical in order to help you succeed.

We certainly believe in microbiome‑based medicine and diagnostics, but we want to take a step back. What are the key advantages of using the microbiome for therapeutics and diagnostics? And what are the most promising applications over the next five years? In other words, why use the microbiome?

Everything is a system. The microbiome is a major part of that system – and it can be modulated, it’s malleable. It’s harder to change your genes; it’s comparatively easy to change your microbiome. Depending on the literature, roughly half the cells in your body are microbial, so the impact is obvious. It’s valuable both as a biomarker and as a therapeutic lever. As a system, the interaction between the microbiome, your genetics, and the metabolites that flow between the two allows the system to work (though that’s an oversimplification). And because you can collect samples non‑invasively, it’s a great substrate for both treatment and diagnosis. That’s why I think it has high potential.

From an applications and therapeutics perspective, there are links to many conditions. The gut-brain axis has good early data – modulating the gut could improve symptoms in autism and Parkinson’s, for example. GI‑related conditions – IBD or IBS, including UC – are strong opportunities because you’re treating at the source. Another area where we’ve seen good data is skin disorders: eczema, atopic dermatitis and related disorders show improvements both via gut modulation (gut-skin axis) and direct skin microbiome modulation. A recent paper in Nature Communications on the Yanomami people in South America reported much higher skin microbiome diversity than in industrialized populations. This suggests that we’ve lost skin biodiversity through industrialization, with implications for both drug and cosmetic applications.

I’m especially excited about drug response’s increasing role as a companion diagnostic. Ultimately, I think a stool test will be part of an annual physical, serving as a disease prediction indicator. Infectious disease is another clear area: there are a couple companies working in this area right now. One is producing blood‑based metagenomic tests that can detect infections missed by PCR, and another is focusing on CSF. Those are more technology plays than microbiome‑profile plays, but the applications are clear.

You’ve mentioned enablers that have made these approaches more feasible – better datasets, lower sequencing costs, and so on. What are the greatest unmet needs or outstanding challenges preventing widespread adoption? What will it take for microbiome profiling to become part of routine care?

More clinical data and more clinical studies. Medicine doesn’t change quickly. You guys had a great post on this recently on the diagnostics side. If you have enough strong clinical data, you can navigate regulation, but generating that data requires funding – it’s a bit of a chicken‑and‑egg problem – you need data so you can get approval and then get money, but you need money to be able to collect good data in the first place, too. The FDA is more open now that some drugs are cleared; there’s a framework to adapt to. But we need data that leads to trials that lead to approvals. Without approval, you don’t get reimbursement, and payers move slowly until you show clinical and economic benefit.

That ties into the competitive landscape as commercialization ramps up. What do you see as the key competitive dynamics, and how do you see them evolving?

I see three primary moats:

First there’s IP. For diagnostics: panels and methods. For drugs: the organisms and their combinations, and what’s protectable there.

Then comes the data – the amount you have and how much you can generate. Highly curated microbiome profiles with robust metadata – ideally EHR‑level clinical data – enables better product development and in‑silico models. Those are two big moats on the competitive landscape side.

A third could be scale and efficiency. That’s the next milestone you should aim for, whether you’re running your business or your trials. Operational excellence, being AI‑first – all in order to execute studies and reach milestones efficiently.

Any final thoughts on trends or developments you have your eye on across microbiome‑based medicine?

We’ve talked a lot about the microbiome specifically, but really we’re trying to understand and improve the system – whether through therapeutics or diagnostics. It’s very important that in the next wave we’re not just looking at microbiome data alone. We need multimodal inputs: genomics, metabolomics, EHR, imaging, and more. That’s how we’ll move forward faster and that’s how we unlock AI’s power – linking across data types to be more expansive and robust will help us reveal relationships we’d miss otherwise. That deeper systems understanding will yield better drugs and better diagnostics.

‍

Thank you so much for a great discussion, Manoj. We appreciate your time and insights.

‍