MeSH Counter: A Tool to Track Biomedical Research Trends

Venture investing in early-stage companies hinges on spotting transformative technologies before they become hyped. To support this, we built the MeSH Counter: an R-based tool that tracks monthly frequencies of specific terms in PubMed, allowing us to visualize trends in the life sciences. Publication trends are not the same as investment signals, yet the MeSH Counter can help efficiently map the research landscape, so investors can prioritize deeper, human-led diligence.

By Dennis Pedri

A major part of the venture capital business is identifying innovative, transformative technologies before they become widely hyped — when valuations are still reasonable, before competitors catch on. Life sciences investors naturally track academic and clinical research closely, looking for signals that may translate into high-impact innovations down the line. To support these efforts, we have developed a tool at V-Bio Ventures which we call the ‘MeSH Counter’ — a way to easily visualize what is trending, stable, or fading in the life sciences.

What is MeSH?

The Medical Subject Headings (MeSH) is the U.S. National Library of Medicine’s method for indexing and organizing biomedical literature in the PubMed database. In practical terms, MeSH terms function as standardized, curated tags assigned to each article. They map synonyms and variant phrasing to a single underlying concept and are organized in a hierarchical tree, from broad categories to increasingly specific topics. This structure makes it possible to expand searches to include related subtopics, or narrow them to a precise niche.

Because MeSH is curated and standardized, it provides a naturally suitable substrate for trend detection. A quick way to spot emerging themes is to review the new MeSH descriptors introduced each year. While the 2025 list is still being finalized, the 311 new descriptors added in 2024 are publicly listed by the National Library of Medicine (NLM). Several of these additions — such as Micropeptides, Enhancer RNAs, and Metabolic Reprogramming — reflect areas of basic biology that have recently gained prominence.

Counting MeSH Terms at Scale

At V-Bio, we wanted a more granular, quantitative view of what is gaining (or losing) attention in academic literature. That is why we built the MeSH Counter.

In short, it is an R-based tool that counts MeSH term occurrences per month and normalizes them to the total number of PubMed publications over the same period. PubMed added more than nine million publications in the last five years, so at first glance this may seem like a daunting task. Fortunately, the National Center for Biotechnology Information (NCBI) provides programmatic access to PubMed via the Entrez Programming Utilities (E‑utilities) API, which we leveraged for our automated data collection.

For our initial implementation, we extracted MeSH terms from the following top-level categories: Anatomy, Organisms, Diseases, Chemicals and Drugs, Analytical, Diagnostic and Therapeutic Techniques and Equipment, Psychiatry and Psychology, Phenomena and Processes, Disciplines and Occupations, Technology, Industry and Agriculture, Information Science, Health Care. We ended up with a total of 36,724 individual terms.

Identifying Trends

Once we had collected the monthly occurrence for each of these MeSH terms, the next challenge was ranking ‘what’s trending?’.

To do this, we used a simple approach. For each term, we fit trend lines across two fixed time spans: a five-year period (giving us its long-term trajectory) and the most recent 12-month period within that time (for its near-term acceleration).

Figure 1: Normalized monthly occurrence of the term ‘Generative Artificial Intelligence’ in PubMed from 2021 to 2025, with trend lines for the total 5-year duration (blue) and last 12 months (red).

From these figures, we computed three metrics for each MeSH term:

• Slope: the coefficient of the fitted line, i.e. the average rate of change of the term’s occurrence over the time period.
• Delta: the net change between the minimum and maximum value in the window, i.e. the overall shift in popularity (less sensitive to short-term ‘wobbles’).
• Momentum: a short-term acceleration score comparing the average normalized frequency in the most recent 12 months to the average in the preceding 12 months, expressed as a relative change (for example, a momentum of 0.25 meant that average for the most recent 12 months is 25% higher than the average of the prior 12 months).

Each metric captures a different measure of a term’s growth: slope emphasizes its average trajectory, delta emphasizes the magnitude of change, and momentum emphasizes its recent acceleration.

Visualizing the Results

To make the outputs easier to interpret, we created heatmaps that plot the normalized occurrence of the selected terms over time. For example, the 200 terms which showed the strongest increase in usage in PubMed over a five-year period.

In these heatmaps, each term’s row is scaled to its own minimum and maximum. This ensures every term spans the full color range, making patterns comparable across terms and highlighting within-term dynamics over time, rather than differences in absolute prevalence across terms.

Figure 2: Heatmap of a selection of terms from the 200 MeSH terms with the strongest five-year growth (by normalized monthly occurrence in PubMed from 2021 to 2025).

To navigate the complex world of early-stage biotech, investors have to do more than analyze financials and clinical milestones — they must reckon with their own biases and cultural tendencies.

Understanding the psychological divide between US and European investors isn’t just a curiosity — it’s a strategic asset. Those who learn to blend the bravery of American venture capital with the wisdom of European finance will not only build stronger portfolios — they’ll also be better positioned to fund future health solutions, to everyone’s benefit.

We also looked at the inverse: which terms have been declining most in usage over the same five-year period.

Figure 3: Heatmap of a selection of terms from the 200 MeSH terms with the strongest five-year decline (by normalized monthly occurrence in PubMed from 2021 to 2025).

Finally, we visualized momentum — each term’s relative change in prevalence from one month to the next — to highlight topics that may be gaining traction in the most recent literature.

Figure 4: Heatmap of a selection of terms from the 200 MeSH terms with the highest momentum (by the difference to the previous month in normalized monthly occurrence in PubMed from 2021 to 2025).

A few key learnings about our MeSH Counter methods have already emerged from our analysis.

• Several topics that the scientific community widely recognizes as important can act as useful ‘positive controls’ for analyzing the trends of newer or lesser-known terms. For example, over the last five years there has been a steady rise in the occurrence of terms such as Metabolism, Drug Resistance, Aging, Tumor Microenvironment, Depression, Cell Death, Neurocognitive Disorders, Mitochondria, Macrophages, and Ferroptosis, but also a similar trend for terms like Computational Biology, Molecular Docking Simulation, Multiomics, and Artificial Intelligence.
• Terms can also be used as benchmarks, or to understand related patterns, for example with the unsurprisingly sharp declines in Covid 19 Testing and Covid 19 Vaccines from 2022 onwards, which correspond with similar declines in Genes and Vaccines.
• Similarly, it’s worth noting when terms may form a part of a larger trend in a particular field. For example, the notable uptick in AI research in the past year can be observed in the increased prevalence of terms such as Generative Artificial Intelligence, Quantum Theory, and Large Language Models. Other broad topic categories showed the same trend, such as terms related to weight loss and drug development.
• As a measure, momentum is inherently noisier than the overall trend and should be interpreted with more caution. That said, it can be useful for surfacing topics that may not yet be on our radar, while still capturing recognized macro-trends — notably, the explosion of AI research.

The MeSH Counter is most powerful if the exercise is repeated every few months, comparing the results from different points in time for more insights and conclusions. However, when using the tool, it is important to bear in mind that some terms move in waves, with sharp spikes and declines over short periods. This is to be expected: researchers respond to pivotal papers, and ‘follow-on’ work often created publications clustered in time. Biases might also be introduced by events like major conferences or special journal issues focusing on specific fields.

This effect was evident when we compared our latest results to a MeSH analysis we conducted three months earlier. In that comparison, the top 200 terms were highly stable, with 89% overlap between the two five-year periods. By contrast, there was far more variability between the most recent 12-month periods (about 20% overlap), and between the momentum rankings. For example, while some terms were present in both analyses (e.g.) Artificial Intelligence, GLP1, Molecular Docking Simulation, High Throughput Screening Assays, and Ferroptosis, other terms were present in the first but were gone from the top 200 terms list just three months later (e.g. Alzheimer’s Disease, Tauopathies, Mood disorders, and Sleep Disorders).

Limitations and the Human Touch

The MeSH Counter has at least two important limitations.

Firstly, there is an indexing lag: MeSH terms are assigned after publication – usually a month or two — so the tool should not be used to interpret trends in the most recent months.
Secondly, publications are not endorsements for investment: an increase in the prevalence of a term or topic does not necessarily imply that there is an investment opportunity or long-term value in the field. Research activity can be driven by many factors, including funding cycles, methodological fads, disease outbreaks, or other transient forces.

As with most tools, the MeSH Counter is useless if not guided by human intelligence. However, when used appropriately, it can efficiently map how the biomedical research landscape is shifting. At V-Bio Ventures, we see it as a tool that strengthens our ongoing scouting efforts for transformative health innovations, by helping us ask better questions and spot patterns earlier.

For investors, the MeSH Counter can provide a complimentary addition to the insights we only glean from deep domain expertise, and the essential human relationships we build with peers, entrepreneurs, and researchers.