The virtual biotech model: advancing a drug with no labs and a team of five

A virtual biotech advances a drug candidate with no laboratories of its own. A small core team owns strategy, scientific judgment, regulatory accountability, and vendor oversight, and contracts out chemistry, manufacturing, toxicology, and clinical operations. The model works when the core team can direct and audit its vendors. It fails when the sponsor forgets that FDA holds the sponsor, not the vendor, responsible.

What is a virtual biotech, and how does it differ from a startup with labs?

A virtual biotech is an asset-centric company that holds the intellectual property, regulatory filings, and decision rights, and buys nearly everything else as a service. Forster and colleagues described virtual pharmaceutical companies as differing from other biotech startups because they "solely comprise experienced teams of managers," with no internal laboratory work [1].

The difference from a lab-based startup is where the fixed cost sits. A lab-based company carries facilities, equipment, and staff whether or not a program is active. A virtual company converts most of that into variable spend tied to specific work orders.

What can a virtual biotech safely outsource?

In my experience, functions that outsource well share three features: the deliverable can be specified precisely, the quality standard is codified, and many qualified vendors compete. Discovery chemistry under a defined plan, GLP toxicology, GMP manufacturing at a contract development and manufacturing organization (CDMO), and clinical site monitoring all fit.

GLP toxicology is the clearest case. 21 CFR 312.23 requires, for each nonclinical study subject to Part 58, a statement that the study complied with good laboratory practice or a brief explanation of why it did not [5]. The standard is written down, and a qualified facility knows what it must deliver.

Clinical operations outsource well for the same reason. FDA's 2013 guidance on risk-based monitoring asks sponsors to focus oversight on the aspects of study conduct most important to subject protection and data quality [6]. A CRO can execute that plan. Deciding what counts as critical is the sponsor's job, and I never hand it off.

What must stay in-house, and why?

The functions that must stay in-house are the ones where the company's judgment is the product. Choosing the indication, interpreting data, deciding whether to advance or stop, owning the regulatory relationship, and managing vendors are not services. If they are outsourced, the company has no reason to exist.

Who is accountable to FDA when the work is done by vendors?

The sponsor. The regulation defines a sponsor as "a person who takes responsibility for and initiates a clinical investigation," and a contract research organization as an independent contractor that assumes one or more sponsor obligations [8]. Assuming an obligation does not relieve the sponsor of it.

The sponsor's general responsibilities under 21 CFR 312.50 include selecting qualified investigators, ensuring proper monitoring, and maintaining an effective IND [9]. Section 312.56 adds that the sponsor "shall monitor the progress of all clinical investigations being conducted under its IND" [10]. A CRO can perform the monitoring. The duty to see it happens is mine.

ICH's revised good clinical practice guideline, E6(R3), issued by FDA as final guidance in September 2025, agrees. Principle 10.1 states that a sponsor may transfer trial-related activities but retains overall responsibility; Principle 10.2 states that responsibility for trial conduct and data integrity "resides with the sponsor" even when work goes to service providers [7].

Manufacturing follows the same logic. FDA's quality agreements guidance states that "no matter who tests the products, the owners' quality units are ultimately responsible for ensuring that the products are manufactured in accordance with CGMP," and a quality agreement does not change that [3]. A five-person company still needs a quality unit, even if that is one person and a consultant.

Does the virtual model actually save money?

It saves overhead, not the science required to reach approval. Wouters and colleagues estimated the median capitalized R&D investment per new drug approved between 2009 and 2018 at $1,141.7 million in 2018 dollars, across 63 products, in figures corrected in 2022 [4]. For antineoplastic and immunomodulating agents, the median was $2,771.6 million [4].

DiMasi and colleagues, studying 106 compounds from 10 firms, estimated out-of-pocket cost per approved compound at $1,395 million and capitalized cost at $2,558 million in 2013 dollars [11]. The studies disagree on magnitude, but both include the cost of failures. A virtual biotech does not escape failure cost; it only avoids paying for idle laboratories meanwhile.

Schuhmacher and colleagues measured R&D efficiency at major companies from 2006 to 2014 at USD 3.2 to 32.3 billion per approved drug, and grouped outsourcing and risk-sharing under cost reduction, one of three categories of response they describe [12]. The virtual model responds to those numbers; it does not cure them.

Where does the virtual model break down?

Tacit knowledge leaves with the vendor. When a CDMO finishes a campaign, the people who know why a step was tuned a certain way move to another client. I require development reports that record reasoning, not just results, and attend technical meetings myself. That coordination cost never appears in a vendor quote.

Dependency risk is real. If a critical-path vendor slips or deprioritizes a small client, the program stalls with no internal capacity to absorb the work. Forster and colleagues noted in 2014 that virtual companies had been described only anecdotally [1], and I know of no rigorous data since on how often vendor failure ends a program. I keep a second qualified vendor on file for each critical function.

Management attention is the hidden cost. Five people cannot oversee twelve vendors with the depth 21 CFR 312.50 implies [9]. DeCorte's criteria for choosing a CRO, including intellectual property safeguards, communication quality, and data integrity, are also what the sponsor must keep checking after signing [13]. The model works only if the core team knows what to check.

How did it work on my own programs?

I have operated in a virtual company that developed first-in-class small-molecule prodrugs alongside a lead first-in-class biologic for a rare pediatric leukemia, with a core team of about five. Chemistry, GLP toxicology, GMP manufacturing, and clinical operations were contracted. Strategy, regulatory authorship, and data interpretation stayed with us.

That biologic reached FDA IND acceptance, received Fast Track designation two months later, and entered first-in-human Phase 1 trials. I oversaw bioanalytical assay development for that trial, including PK/PD, anti-drug antibody, biomarker, potency, and safety assays, while contract laboratories ran them.

The model could not make the science faster. GLP toxicology and GMP manufacturing took the time they take regardless of who signed the invoice. It did let a very small team carry two first-in-class programs, in two modalities, to decision points a lab-based company of the same size could not have afforded.

This article is part of the Product Development Services practice at Hamilton Atlas Group — see the full capability list there.

References

1. Forster SP, Stegmaier J, Spycher R, Seeger S. Virtual pharmaceutical companies: collaborating flexibly in pharmaceutical development. Drug Discovery Today. 2014;19(3):348-355.
2. Legal Information Institute, Cornell Law School. 21 CFR § 312.52 - Transfer of obligations to a contract research organization. Code of Federal Regulations, 2025.
3. U.S. Food and Drug Administration. Contract Manufacturing Arrangements for Drugs: Quality Agreements Guidance for Industry. FDA, 2016.
4. Wouters OJ, McKee M, Luyten J. Estimated Research and Development Investment Needed to Bring a New Medicine to Market, 2009-2018. JAMA. 2020;323(9):844-853.
5. Legal Information Institute, Cornell Law School. 21 CFR § 312.23 - IND content and format. Code of Federal Regulations, 2025.
6. U.S. Food and Drug Administration. Oversight of Clinical Investigations — A Risk-Based Approach to Monitoring Guidance for Industry. FDA, 2013.
7. U.S. Food and Drug Administration. E6(R3) Good Clinical Practice (GCP). FDA, 2025.
8. Legal Information Institute, Cornell Law School. 21 CFR § 312.3 - Definitions and interpretations. Code of Federal Regulations, 2025.
9. Legal Information Institute, Cornell Law School. 21 CFR § 312.50 - General responsibilities of sponsors. Code of Federal Regulations, 2025.
10. Legal Information Institute, Cornell Law School. 21 CFR § 312.56 - Review of ongoing investigations. Code of Federal Regulations, 2025.
11. DiMasi JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry: New estimates of R&D costs. Journal of Health Economics. 2016;47:20-33.
12. Schuhmacher A, Gassmann O, Hinder M. Changing R&D models in research-based pharmaceutical companies. Journal of Translational Medicine. 2016;14:105.
13. DeCorte BL. Evolving Outsourcing Landscape in Pharma R&D: Different Collaborative Models and Factors To Consider When Choosing a Contract Research Organization. Journal of Medicinal Chemistry. 2020;63(20):11362-11367.