What 18,000 Preprints Reveal About How Scientists Collaborate

By shrijeetverma13 · July 28, 2026

A detailed analysis of bioRxiv submissions shows neuroscience dominates the platform, Stanford leads in output, and most fields rarely publish their…

Neuroscience is by far the most dominant research category on bioRxiv, accounting for 3,679 preprints (20.44%) out of 17,997 total submissions. It's followed by bioinformatics (1,942 preprints, 10.79%), microbiology (1,627 preprints, 9.04%), biochemistry (887 preprints, 4.93%), and cell biology (877 preprints, 4.87%). Together, these top 5 categories make up just over half (50.07%) of all preprints, while the remaining submissions are spread across 20 other categories.

Stanford University produces the most preprints in this corpus with 148, followed by University of Michigan (101) and University of Pennsylvania (99). A bar chart ranks the top 15 corresponding institutions by preprint count.

Average abstract length does vary noticeably across the 25 research categories analyzed, ranging from about 1,321 to 1,974 characters. The overall average across categories is roughly 1,699 characters. Pathology has the longest abstracts on average (~1,974 characters, based on 126 papers), while paleontology has the shortest (~1,321 characters, based on 22 papers).

Evolutionary biology shows the highest average version number (1.547), indicating it undergoes the most frequent revisions among all categories. It's followed closely by pathology (1.532), neuroscience (1.503), animal behavior and cognition (1.5), and bioinformatics (1.473). Neuroscience stands out for having the largest volume of preprints (3,679) while still maintaining a high revision rate, and bioinformatics has the highest single max version (12), suggesting some papers in that field get revised extensively.

Preprint submissions were tracked weekly from April 6, 2026 to July 6, 2026, spanning 14 weeks. On average, about 1,286 preprints were submitted each week. Submissions peaked during the week of May 18, 2026 with 1,769 preprints, while the lowest week was July 6, 2026 with 501 submissions (likely a partial week). A line chart visualizes this weekly trend over the full date range.

Across all preprint categories, only about 2.86% eventually reach published status. This rate varies by field: plant biology leads with 4.36% published, followed by synthetic biology (4.14%) and cell biology (4.1%). On the low end, zoology and paleontology show 0% published among their preprints, while genomics sits at just 1.54%. A bar chart breaks down the published percentage for the top 20 categories, making it easy to compare fields at a glance.

The average number of authors per paper has been decreasing over the dataset's time span, going from 7.95 authors in April 2026 down to 7.59 authors by July 2026. The overall average across the whole period was 7.91 authors per paper.

On average, papers have about 8.02 authors each. A bar chart was generated comparing the top and bottom research categories by average author count, showing which fields tend to favor large collaborative teams versus smaller or solo efforts. The visualization highlights the 10 categories with the highest average author counts alongside the 10 with the lowest, making it easy to see the contrast in collaboration styles across research areas.

By comparing each institution's peak weekly submission count to its own average weekly rate, the analysis flagged 168 institutions showing unusually concentrated bursts of paper submissions. The strongest anomalies include University of California, Davis (8 papers in one week vs. a 2.4 average, a 3.4x spike) and University of Glasgow (9 papers vs. 2.9 average, 3.1x spike). Other notable cases include University of Alabama at Birmingham, University of New South Wales, Kyoto University, University of Tuebingen, Utrecht University, and Indian Institute of Technology Kharagpur, all showing peak weeks at least 2.8x their normal submission pace.

Papers average about 8 authors each, with a standard deviation of roughly 6. Using the 3-sigma method, any paper with more than 26 authors is considered an outlier — and 309 papers in the dataset meet that criteria. A histogram shows the overall distribution of author counts with the outlier threshold marked, and a bar chart highlights the top outlier papers by author count along with their categories, making it easy to see which fields tend to have very large collaborative teams.