The long road to NM102: a case study in scientific persistence
Research data bite: 19.
Key takeaways
Breakthroughs are built over decades of sustained research. To truly reflect scientific progress, we need lineage-aware, cumulative metrics that capture the long-term intellectual labor and persistence leading up to pivotal discoveries—not just the impact of a single article.
NM102 is a paradigm shift: an anti-virulence approach that avoids collateral damage to the microbiome and offers a new weapon against antibiotic resistance; it could mark a turning point as significant as the discovery of antibiotics.
The other day, one of my favourite podcasts, Avec science, went through a case study that I found really interesting; NM102, une nouvelle piste contre l’antibiorésistance (NM102, a new clue against antibioresistance). Nalini Rama Rao, last author on the publication, and who was interviewed explained she worked for a couple of decades on Mutation Frequency Decline (Mfd) proteins. Her work led to the discovery of NM102, a molecule that does not kill bacteria outright but makes them more vulnerable to the immune system. It works during infection, without harming the helpful bacteria that keep our microbiome in balance.
This is a significant step forward; no entirely new class of antibiotics has been discovered since 1987, and the last one approved by the FDA was in 2003. With antimicrobial resistance on track to become one of the leading causes of death worldwide, a compound like NM102 could help extend the effectiveness of existing treatments. It is not just a new drug; it is a different way of thinking about how we fight infections.. before infections kill more than cancer in 2050.
Timeline
So let’s start with this new publication, out on 28 April 2025. Searching by title through Dimensions’ webapp, we identify two publications: the Nature Communications, as well as the preprint, posted in biorXiv in January 2024, nearly 18 months earlier than the peer reviewed article.
Crossref metadata indicates that the Nature Communications article was:
posted on 23rd August 2024,
accepted on 18 March 2025 (7 months review),
published online on 28 April 2025.
The gap between biorxiv publication and the Nature communications article suggests either that modifications were made by the research team before submission to Nature Communications, or that the article was submitted somewhere else and rejected; 6 months lost of the researchers.
Citations
The Nature Communications article does not currently have any citations, which would be a surprise at this point, less than a month after publication. However, the 18-month old preprint does have a couple of citations, which it turns out are from what seems to be the same publication, but in different languages (English and German; still a visible footprint for German chemistry!). Some may argue these citations should have been merged, after all it is only one set of authors that thought of citing it; however another view is that the translation increases the visibility and therefore highlights more impact.
Now, should the preprint transfer its citations to the postprint? Although it seems obvious in cases where the publications is identical, it is actually not a given that these would be identical; therefore this is a call to be made by the analyst, rather than the data provider. Also, there has been no modification of the record of the preprint to indicate that it had indeed resulted in a publication (it is sometimes indicated at the preprint server level), so it is only a high suspicion based on the title, the authors, and abstract.. but I haven’t made a comparison of the full text and it is not trackable through machines.
Attention
Where the age of the preprint gave it an advantage to accumulate citations (a measure of scholarly attention), the reputation of journal Nature Communications likely gave it an attention advantage. In the first four days, the article had already accumulated 10 news outlet attention, two blogs, and more mentions on X and Bluesky than its preprint.
Below you can see how the new timelines on Altmetric illustrates the rapid growth followed by a plateau (for now..).
The Nature Communications article does not include attention from podcasts, otherwise this would have added to its score.
Impact of the overall research
Moving beyond article-level metrics
Traditional impact measures focus on the success of individual publications. But what about the accumulated effort behind a breakthrough? In this case, the researchers have worked on the same line of inquiry for over two decades. A single high-profile article cannot capture the persistence, intellectual scaffolding, or collaborative dynamics that made it possible.
Reconstructing a lineage
The following chart traces the intellectual lineage behind the Nature Communications publication on NM102. Each dot represents a citation involving overlapping concepts, linking the cited paper’s publication year (vertical axis) to the year it was cited (horizontal axis). Dots are coloured to indicate whether the citation is a self-citation (a co-author of the breakthrough publication was also an author on the cited paper) or an external citation. Together, these links reveal how a long arc of prior work, sometimes authored a few decades earlier, was recursively built upon by the same research team.
Rethinking self-citation
While self-citation is often viewed with skepticism, in this context I believe that it reflects genuine intellectual continuity. The repeated reuse of conceptually aligned, self-authored work suggests a deliberate and evolving research agenda, not opportunistic citation padding. The thematic cohesion of these papers, measured by shared concepts, adds semantic weight to the structural links.
Conclusion
Rather than focusing only on the latest publication’s immediate performance, this lineage chart foregrounds the cumulative labor behind it. It invites a broader perspective: one that values the trajectory of a research group, the networks of influence they cultivate, and the knowledge ecosystems they sustain. Whether NM102 proves clinically successful or not, its emergence is a visible signal of something deeper; a career-spanning commitment that standard metrics rarely capture.
Code
-- STEP 1: Breakthrough paper
WITH breakthrough_paper AS (
SELECT *
FROM `dimensions-ai.data_analytics.publications`
WHERE id = 'pub.1168201811'
),
-- STEP 2: Authors of the breakthrough paper
breakthrough_authors AS (
SELECT a.researcher_id
FROM breakthrough_paper,
UNNEST(authors) AS a
WHERE a.researcher_id IS NOT NULL
),
-- STEP 3: Concepts of the breakthrough paper
breakthrough_concepts AS (
SELECT DISTINCT c.concept
FROM breakthrough_paper,
UNNEST(concepts) AS c
),
-- STEP 4: Papers with any concept match
concept_matched_papers AS (
SELECT DISTINCT p.id AS cited_pub
FROM `dimensions-ai.data_analytics.publications` p,
UNNEST(p.concepts) AS c
JOIN breakthrough_concepts bc ON c.concept = bc.concept
),
-- STEP 5: First-degree citations
first_degree AS (
SELECT DISTINCT
bp.id AS citing_pub,
r.id AS cited_pub,
1 AS level,
bp.year AS citing_year,
r.year AS cited_year,
r.metrics.times_cited AS cited_citation_count,
r.altmetrics.score AS cited_altmetrics_score,
r.title.preferred AS cited_title,
r.journal.title AS cited_journal,
ARRAY_LENGTH(r.research_orgs) AS cited_research_org_count
FROM breakthrough_paper bp
CROSS JOIN UNNEST(bp.reference_ids) AS ref_id
JOIN `dimensions-ai.data_analytics.publications` r ON r.id = ref_id
JOIN concept_matched_papers cmp ON r.id = cmp.cited_pub
),
-- STEP 6: Second-degree citations
second_degree_refs AS (
SELECT cited_pub FROM first_degree
),
second_degree AS (
SELECT DISTINCT
p.id AS citing_pub,
r.id AS cited_pub,
2 AS level,
p.year AS citing_year,
r.year AS cited_year,
r.metrics.times_cited AS cited_citation_count,
r.altmetrics.score AS cited_altmetrics_score,
r.title.preferred AS cited_title,
r.journal.title AS cited_journal,
ARRAY_LENGTH(r.research_orgs) AS cited_research_org_count
FROM `dimensions-ai.data_analytics.publications` p
JOIN second_degree_refs s ON p.id = s.cited_pub
CROSS JOIN UNNEST(p.reference_ids) AS ref_id
JOIN `dimensions-ai.data_analytics.publications` r ON r.id = ref_id
JOIN concept_matched_papers cmp ON r.id = cmp.cited_pub
),
-- STEP 7: Author roles on cited papers
cited_paper_authors AS (
SELECT
p.id AS cited_pub,
a.researcher_id,
CASE
WHEN pos = 0 THEN 'first'
WHEN pos = ARRAY_LENGTH(p.authors) - 1 THEN 'last'
ELSE 'middle'
END AS author_position
FROM `dimensions-ai.data_analytics.publications` p,
UNNEST(p.authors) AS a WITH OFFSET pos
),
-- STEP 8: All citations (first + second degree)
all_citations AS (
SELECT * FROM first_degree
UNION ALL
SELECT * FROM second_degree
),
-- STEP 9: Join all citations with possible author roles (left join)
citations_with_roles AS (
SELECT
ac.*,
ca.researcher_id,
ca.author_position
FROM all_citations ac
LEFT JOIN cited_paper_authors ca ON ac.cited_pub = ca.cited_pub
),
-- STEP 10: Flag self-citations and filter only for breakthrough authors
final_with_flags AS (
SELECT
cwr.*,
CASE
WHEN ba.researcher_id IS NOT NULL THEN TRUE
ELSE FALSE
END AS self_citation
FROM citations_with_roles cwr
LEFT JOIN breakthrough_authors ba
ON cwr.researcher_id = ba.researcher_id
)
-- FINAL SELECT
SELECT
researcher_id,
author_position,
self_citation,
citing_pub,
cited_pub,
level,
citing_year,
cited_year,
cited_citation_count,
cited_altmetrics_score,
cited_title,
cited_journal,
cited_research_org_count
FROM final_with_flags
ORDER BY cited_year, self_citation DESC