Do natural products still matter for drug discovery? Few questions in this field generate as much heat and as little data. Newman and Cragg have spent four decades answering it with bookkeeping: five prior reviews in this journal (1997, 2003, 2007, 2012, and 2016) and now a sixth, published open access in 2020 in a special issue honoring Jon Clardy, classifying every therapeutic agent approved worldwide by where its structure came from. It is the closest thing the field has to an audited answer, and it rewards a close reading because its categories — not just its counts — teach a way of thinking about what natural products actually contribute.

The paper at a glance

The review classifies all 1881 therapeutic agents first approved anywhere in the world between 1 January 1981 and 30 September 2019 — of which 1394 are small molecules and 346 are biologicals — assigning each a source code recording whether the structure is natural, synthetic, or honestly somewhere in between. Antitumor drugs are tracked over the longer frame from 1946, the era of the nitrogen mustards, to the same 2019 cutoff. Only the first approval of any drug is counted; biosimilars and added indications are not. Every compound’s name, disease, year, and source code is tabulated in the Supporting Information, assembled from Annual Reports in Medicinal Chemistry and its successor publications, Drug News and Perspectives, Drugs of Today, and the Clarivate Integrity database. This edition also introduces radar-plot graphics and splits the anticancer tables so that the 1946–2019 and 1981–2019 frames are never silently mixed.

A taxonomy honest about mixed parentage

The analytical core of the paper is the coding scheme, and it is more subtle than the headline percentages suggest. N is an unaltered natural product. NB is a natural product botanical, one of the “defined mixtures” the FDA now recognizes as drug entities. ND is a natural product derivative, typically semisynthetic. S is a totally synthetic drug. The honest innovation is the pair of hybrid codes: S*, a synthetic drug whose pharmacophore came from a natural product, and the /NM suffix, marking a natural product mimic — a synthetic compound that only exists because knowledge of a natural product’s structure or mechanism showed what to build. A binary natural-versus-synthetic tally would hide exactly the categories through which most of nature’s influence flows; S* and /NM make chemical inspiration visible as something countable. When the paper writes “N*” it means the natural cluster N + NB + ND.

The headline numbers

For anticancer small molecules the time frames are kept explicit. From 1946 to 1980, 40 of the 75 approved small molecules (53.3%) are N or ND. From 1981 to September 2019 there were 247 anticancer new chemical entities in total, of which 185 (75%) are small molecules, and the breakdown by source is striking: N 18 (9.7%), NB 1 (0.5%), ND 43 (23.2%), S* 13 (7.0%), S*/NM 45 (24%), S/NM 36 (19.5%) — and S, totally synthetic with no natural lineage, only 29, or 15.7%. The natural cluster N* accounts for 62 of the 185 (33.5%); add the pharmacophore-borrowers and mimics and 120 of 185 anticancer small molecules, 64.9%, trace back to a natural product somewhere in their lineage. The anti-infective areas remain, in the authors’ words, dependent on natural products and their structures, while Table 2 of the review names disease areas where no approved drug derives from a natural product at all.

What the census quietly argues

Two findings carry the argument beyond percentages. The first is the combinatorial verdict: across the entire 39-year frame, the authors can identify only two approved drugs that originated as de novo combinatorial-chemistry compounds — one of them, they note, a little speculative — plus a single drug, approved in 2012, developed through fragment-binding methodology. The technology that was supposed to replace nature as a source of starting points delivered, by this ledger, almost nothing. The second is the paradox of supply: natural products keep feeding the pharmacopeia in spite of what the authors describe as the greatly reduced level of natural product-based drug discovery programs in major pharmaceutical houses. The review closes by looking forward rather than back, tabulating the antibody–drug conjugates then in phase II and III clinical trials (as of 27 December 2019) — agents whose warheads, from the calicheamicin of gemtuzumab ozogamicin to the dolastatin-derived monomethyl auristatins of the vedotin conjugates, are natural products or close derivatives repurposed as payloads.

Limitations and open questions

A close reading should hold onto what a census cannot say:

  • Approvals are not impact. Counting first approvals treats a transformative drug and a marginal one identically, and the ledger records nothing about attrition — how many natural product programs failed for each approval — so it answers “where did approved structures come from,” not “which discovery strategy pays best.”
  • The boundaries are judgment calls. Whether a synthetic drug earns S*, an /NM suffix, or neither depends on deciding that its pharmacophore came from a natural product. The judgments are defensible and auditable — every compound’s code is published — but they are expert judgments, hand-curated from secondary literature rather than produced by a reproducible pipeline, and a reader cannot re-derive them from the paper alone.
  • The ledger closes in September 2019. The antibody–drug conjugate pipeline the authors describe with admitted optimism has, in the years since, partly turned into approvals. Readers in the 2020s should treat these tables as a baseline awaiting the next edition, not as a current verdict.
  • Absence is ambiguous. The disease areas with no natural product-derived drugs could reflect biology, screening history, or decades of fashion in what got screened; the census records the pattern but cannot separate the causes.

What to take from it

For a practicing natural product chemist, three things transfer. First, the taxonomy itself: the next time someone claims that natural products contribute some percentage of all drugs, ask which categories they are counting — N alone is a small number, N + ND is the chemistry claim, and adding S* and /NM makes it the inspiration claim — and cite the figure with its time frame attached, because the 64.9% anticancer number belongs specifically to small molecules approved between January 1981 and September 2019. Second, where the value path actually runs: unaltered natural products are a minority even in the friendliest disease areas, and the dominant routes are derivatives and mimics, which means an isolated molecule’s most likely contribution is as a template — a pharmacophore with evolution-tested biological relevance — rather than as the drug itself. Third, the deepest lesson is the one this publication exists to examine: natural structures carry information shaped by biological function, and four decades of synthetic alternatives have proved far better at exploiting that information than at replacing it. The question to ask of any new isolate is not only “is it active?” but also “what does its structure know that a screening library does not?”