Peptide basics

Publication counts are not evidence: how peptide literature totals mislead

This page is about how counts are produced. It does not evaluate any compound, and a large or small literature is not an argument for or against anything here.

A count is the answer to a question

PubMed does not store a number of papers about a compound. It stores records, and it returns however many match the query you typed, under the interpretation it applies to that query. Change the query and the number changes, sometimes by two orders of magnitude, without a single record having been added or removed.¹²

Two behaviours do most of the damage. A bare term searches all fields, including author names, journal titles, affiliations and indexing terms, rather than only what a paper is about. And PubMed expands terms automatically, mapping what you typed onto related vocabulary, which is helpful for a clinician and catastrophic for a count.¹²

The clearest case: MGF

Searching PubMed for MGF, unrestricted, returns 2,001 records. That figure appears on this site's data for mechano growth factor, and taken at face value it would make MGF one of the best-documented compounds in the directory.¹³

Sort the same search by relevance and read the top results. They include African swine fever virus MGF_110-3L and MGF_110-4L, African swine fever virus MGF-505-7R, and African swine fever virus MGF 360-2L — a family of viral multigene-family genes that has nothing to do with muscle. Also present: a magnesium fluoride material described as MgF2:Mn2+, and a polyp-segmentation neural network called MGF-net.¹⁴

The registry side is unambiguous in the other direction. A ClinicalTrials.gov intervention search for MGF returns over a hundred studies, and not one names mechano growth factor as an intervention. A compound can carry a four-figure citation count and an empty human-trial record simultaneously, because the two numbers are answering different questions about different things.⁵

The second case: IGF-1 DES

Unrestricted, IGF-1 DES returns 723 records. Restricted to title or abstract with the [tiab] tag, it returns six. The gap is not noise; it is the entire IGF-1 literature being pulled in by term expansion, which is why the relevance-ranked results for the loose search include general reviews of acromegaly and of IGF-1 in the cardiovascular system.¹

The six that survive the restriction are the real literature, and they are mostly old. One of the more cited is a 1996 review describing Des(1-3)IGF-I as a truncated form of insulin-like growth factor-I — a molecular characterisation, in a biochemistry journal, thirty years ago.⁸

Why this site publishes both numbers

The market data on this site carries an unrestricted term count on the compound record and a strict title-and-abstract count, by year, in the research-activity series. They disagree, sometimes enormously, and both are kept deliberately. The loose number is what a reader who searches naively will see; the strict number is what the compound actually has. Publishing only the strict one would be cleaner and would hide the trap.¹²

The rule that follows for reading anyone's citation count, including ours: a number without its query is not interpretable. If a claim cites a literature size, the honest form states the database, the exact search string, the field restriction, the date range, and the date the search was run.²

Even a clean count is not evidence

Suppose the matching is perfect and the number is real. It still does not tell you what kind of research it counts. The 2025 systematic review of the BPC-157 literature screened 544 articles and included 36 studies, of which 35 were preclinical and one clinical — a compound with a substantial, correctly matched literature and almost no human component.⁶

And the literature being counted is skewed before anyone counts it. In an analysis of 525 publications from animal stroke studies, only about 2 per cent reported no significant effect on the main outcome, and adjusting for publication bias cut pooled reported efficacy from 31.3 per cent to 23.8 per cent. Volume and favourability rise together for reasons that have nothing to do with the substance under study.⁷

What to do with a number you are shown

1. Ask for the query. Database, search string, fields, date range, and run date. A count without them cannot be reproduced or compared.
2. Re-run it with the name in quotation marks and [tiab] appended. If the number collapses, the original was matching something else.
3. Read five titles from the relevance-ranked results. Acronym collisions are obvious within seconds and invisible in a total.
4. Compare equal-length date windows rather than lifetime totals, so database growth does not masquerade as research interest.
5. Then check the registry separately. Literature size and registered human trials are independent facts, and the second is the one most claims are actually about.¹²⁵