Hello Reader, I hope you have read some interesting studies lately.
Why? Because staying up to date on research can reduce our footprints!
Farley et al. have published a preprint in which they tried to estimate the carbon footprint of science that fails to do that…
Their result: about 38 tons of CO2e per study. But is that realistic?
Today's Lesson: The Footprint of Science
Discussing the impact of unnecessary studies
Number Of The Day
According to Retraction Watch, Springer Nature retracted 2923 articles in 2024. Of course, each of these articles not only wasted resources but may also have misinformed subsequent scientific work. It's important to note that with over 3000 journals, Springer Nature published more than 482000 articles that year. Still, every retraction represents a failure of the peer-review system. But what about papers that go unnoticed or involve unnecessary work? Let's find out:
2923
If Science Misses To Self-Correct
Farley et al. recently uploaded a preprint, examining the footprint of unnecessary 5-HTTLPR related research.
The graphic depicts the synaptic cleft, where serotonin (5-HT) is released and reabsorbed. It was created by Warden and Haney, who state: "Blockade of the serotonin transporter increases serotonin concentrations within the synapse in the interneuronal (synaptic) space." [Reproduced from Future Rheumatology (2007); 2(2):213–22 with permission of Future Medicine Ltd.]
It is an interesting case: a gene polymorphism of the serotonin transporter 5-HTT (known as 5-HTTLPR), was linked to anxiety and depression in the late 1990s.
However, large-scale studies published in 2005 and 2009 already suggested there was no strong association with depression.
Still, research on 5-HTTLPR continued for years.
So, how do you calculate the footprint of scientific studies and what did they find?
Studies
At first, Farley et al. estimated the number of studies conducted - just as a meta-analysis would do.
They found 1183 publications from 1996 to July 2024, with 779 published after 2009 (when it should have been obvious that any additional research is unnecessary).
Interestingly, study numbers grew just as Farley et al. argue evidence for a missing link emerged. Given that projects often take 4–5 years—and longer with follow-ups—this rise likely reflects momentum despite contradictory data. Still, interest clearly declined after 2016, suggesting researchers noticed the evidence but didn't halt ongoing work.
Of course, they also pulled metadata: article type, number of authors, etc., both for later estimations and to exclude reviews.
Genotyping
To estimate the footprint of the genotyping necessary for the studies, they extrapolated the number of assays conducted and multiplied it by the average amount of plastic waste involved.
Estimated emissions: 616.3 tons CO2e for all studies (893,165 assays); of these, 405.8 tons CO2e were linked to studies deemed “unnecessary” (from 2010-2024).
Commuting Carbon Footprint
Then, we have to account for the travel of researchers to their workplace. They assumed an average commute to work of 21 km and that researchers would work 0.5–3 years on their studies.
Estimated Emissions: 2308–13 846 tons CO2e for all studies, and 1537–9219 tons CO2e for 2010–2024.
They did not include patients commuting to study centers, but the estimated footprint would be approximately 1992 tons CO2e for all studies and 131 tons CO2e for 2010–2024.
Conference Travel
Based on a study on travel distance and frequency based on scientist's seniority, Farley et al. estimated:
5599 tons CO2e for all studies and 3687 tons CO2e for 2010–2024.
Laboratory Energy Use
Here, Farley et al. used a straightforward approach:
Farley et al.’s assumption on energy consumption: Please note that there are institutes and facilities running without (or with less) gas, which often reduces their overall footprint. Indeed, HVAC typically makes up 50-60% of all energy consumption. “Else” probably includes servers, computers, lighting, etc.
They used the average number of authors (assuming 1 PI + 6 researchers), the average lab space (at 111.5 m²), and average energy consumption (based on the S-Lab audit) to estimate 96.9 GWh for all studies and 63.8 GWh for those from 2009-2024.
These 96.9 and 63.8 GWh translate roughly to 32 791 and 21 590 tons CO2e, respectively.
While there might be discussions on how one accounts for laboratory energy consumption, as it is hard to “split” factors like HVAC, or how big the actual impact of genotyping is (probably strongly underestimated since the impacts of chemicals are hard to assess and Scope 3 emissions are a topic of discussion), we can clearly see that impacts since 2010 have been much higher, hinting at the faster pace of research.
Applying The Knowledge
Taken together, Farley et al. estimated a total carbon footprint of 30068 tons CO2e for studies published from 2010 to 2024.
That’s 38 tons of CO2e per study or 5 tons per contributing scientist!
Now, I’d argue we can safely assume a 10-fold variation in either direction, which means 3 to 300 tons per study is a realistic range.
Click to enlarge. On the left you see a graph from Lesch et al. showing differential 3H15-HT uptake in A), and ligand binding in B) in human lymphoblast cell lines based on polymorphism-genotypes I/l (n = 4), I/s (n = 3), and s/s (n =3) before and after treatment with PMA or forskolin. In essence: the short polymorphism makes a difference in cells. On the left, we see a graph from Risch et al. with the boxes and lines indicate the odds ratios (ORs) and their 95% confidence intervals (CIs) on a log scale for each study. The size of the box indicates the relative weight of each estimate. Their conclusion: no effect of the polymorphism on depression. In the end, the question is, if we see a clear biochemical effect on a cellular level, when do we stop if we don’t find one on the human level. Remember, it might just be the way we assess psychological measurements. Of note, initial studies such as those by Lesch et al. focused on psychological traits such as neuroticism, not depression.
Why This Variation?
Their approach is straightforward: find an average impact/factor and multiply it by the numbers relevant to your case. Of course, this is where a long list of assumptions begins:
Generalizations: commuting distances, number of genotyping assays, which conversion factors to choose
Omissions: failed experiments, unpublished studies due to missing significance, chemical impacts
Accounting overlap: Would lab impacts like heating and conference travel have existed if scientists worked on other topics?
Although the study serves a crucial function, we have to remember it is a preprint. That means:
I missed a lot of methodological detail in their paper. Some calculations, like the footprint of genotyping assay, were pretty unclear. Moreover, their literature search strategy (e.g., looking for “5-HTT” instead of “5-HTTLPR”) seems simply erroneous to me.
Please note that Farley et al. included all research (excluding reviews) on 5-HTT (the receptor) and depression or anxiety, including studies on other mutations or research questions, not just those on the polymorphism. The picture demonstrates what I mean, yielding 1139 studies including reviews. A tip for you: PubMed handles Boolean operators differently from Google. To correctly search for studies linking 5-HTTLPR with either condition, we have to use:"5-HTTLPR" AND (Depression OR Anxiety) which seems a bit unintuitive.
Also, presenting one final footprint number (and suggesting it is probably an underestimate) isn’t exactly best practice, in my view.
However, as the authors rightfully emphasize, transparent communication of results is more crucial than ever given the growing speed and amount of science.
Furthermore, placing more emphasis on literature review - and less on publishing pressure - will be a crucial step to reduce environmental impacts AND enhance scientific robustness.
Upcoming Lesson:
Sustainability Education
How We Feel Today
References
Farley, M. et al., The carbon footprint of science when it fails to self-correct. bioRxiv, 2025. doi:10.1101/2025.04.18.649468.
Warden, S.J. et al., Skeletal effects of serotonin (5-hydroxytryptamine) transporter inhibition: evidence from in vitro and animal-based studies. Journal of Musculoskeletal and Neuronal Interactions, 2008, 8(2), 121–132. PMID: 18622081; PMCID: PMC4155922.
Ciers, J. et al., Carbon Footprint of Academic Air Travel: A Case Study in Switzerland. Sustainability, 2019, 11, 80. https://doi.org/10.3390/su11010080
Lesch, K.P. et al., Association of anxiety-related traits with a polymorphism in the serotonin transporter gene regulatory region. Science, 1996, 274(5292), 1527–1531. doi:10.1126/science.274.5292.1527.
Risch, N. et al., Interaction between the serotonin transporter gene (5-HTTLPR), stressful life events, and risk of depression: a meta-analysis. JAMA, 2009, 301(23), 2462–2471. doi:10.1001/jama.2009.878. Erratum in: JAMA, 2009, 302(5), 492.
If you have a wish or a question, feel free to reply to this Email. Otherwise, wish you a beautiful week! See you again on the 31st : )
Edited by Patrick Penndorf Connection@ReAdvance.com Lutherstraße 159, 07743, Jena, Thuringia, Germany Data Protection & Impressum If you think we do a bad job: Unsubscribe
Personal Note From Patrick, The Editor Hi Reader, I’ve heard from many scientists that teaching is the most fulfilling work they’ve ever done. Teaching students and new colleagues is a powerful tool that is all too often overlooked. Of note, we previously talked about how to convince others, including some common mistakes to avoid. Therefore, we will explore how to effectively embed sustainability in your teaching in today's lesson: Today's Lesson: A Strategy For Your Teaching How we can...
Personal Note From Patrick, The Editor Hi Reader, last week I posted about an essential step to make your lab greener: Quantification. I thought this post might be of interest to you - also because I share some of my experience in the comments. And while we often overlook the benefits of apparently trivial actions like quantification, the same can be said about innovations. A good example is trypsin alternatives — let me explain the issue: Today's Lesson: Trypsin Alternatives How enzymes can...
Personal Note From Patrick, The Editor Hi Reader, hope you are doing well! In a previous lesson, we looked at how to account for the energy consumption of your lab equipment. However, there are three reasons why we’re interested in assessing the exact energy use of a specific piece of equipment. Therefore, today we’ll cover what these are and how you can successfully measure them without becoming too technical: Today's Lesson: Measuring Energy Consumption How to quantify electricity use...