Social network graphs are pretty, but they’re not the only way we can try to visualized cross-institutional research collaborations. Here’s a geographic view of some of the institutions that UCSF dentistry researchers have co-authored with over the course of 2013.
Looking at cross-institutional co-authorship networks is a useful way of seeing not only who we work with, but also where there may be gaps of interest.
I first looked at dentistry-related publications by UCSF researchers published in 2013, breaking out the institutions we co-authored with. And there we are, sitting pretty in the center of our universe, collaborating with major institutions in the US, Korea, Australia, Italy, Denmark, and more.
(Details: Institution node sizes indicate the total volume of dentistry-related articles published. Connecting line widths indicate the number of articles co-authored between two institutions. Distance between nodes indicates the tightness of co-authorship networks, and different sets of node colors help distinguish groups of institutions whose researchers frequently co-author together. Of 462 institutions that collaborated with UCSF researchers, we’re showing only 91 that had 10 or more cross-institutional articles in that time.)
View full-size visualization (PDF)
Then I looked at the total universe of dentistry-related publications published in 2013 (see below). Notice a difference? I have to admit that it took me a while to find UCSF in the mess of dots. (If you look at the full-size view, we’re in the medium blue section, next to the pinks.) Of course this says more about the sheer volume of research being published by universities all over the world, than about any lack of cross-institutionally collaborative spirit on our part; in fact I hid over 80% of the institutions in the first image to keep it readable, which accounts for a a good chunk of the difference. But the sheer weight of institutions from Europe, East Asia, and Latin America in this second image that aren’t there in the first is intriguing, and something I’m going to try digging into.
(Details: Institution node sizes indicate the total volume of dentistry-related articles published. Connecting line widths indicate the number of articles co-authored between two institutions. Distance between nodes indicates the tightness of co-authorship networks, and different sets of node colors help distinguish groups of institutions whose researchers frequently co-author together. Of 2,575 institutions that we found, we’re showing only 374 that had 10 or more cross-institutional articles in that time.)
View full-size visualization (PDF)
(And yes, I realize fully well that I’m probably looking at the wrong things here, privileging increasing the count of cross-institutional collaborations as an end in itself, avoiding any consideration of research quality, and giving greater visual weight to institutions that publish more, regardless of the size of the institution or the quality of work. Pretty pictures lie can hide lots of flaws. I hope you’ll bear with me as I publicly iterate through these topics, step by step, hopefully getting just a little bit less dumb every time.)
Additional uninteresting details: I searched Web of Science for dentistry-related articles published in 2013 (i.e. from January 1-December 5, 2013). I began by running a search for any articles published in 2013 matching a number of dentistry-related keywords (dental, dentistry, electrogalvanism, endodontics, jaw relation record, mouth rehabilitation, odontometry, oral, orthodontics, periodontics, prosthodontics, teeth, tooth), then filtered only those that matched the “DENTISTRY ORAL SURGERY MEDICINE” Web of Science category.
I was looking to dig into some examples of collaboration patterns in different research areas, when I realized I didn’t even know the basics — what do UCSF researchers actually research?
UCSF Profiles uses PubMed data to extract MeSH keywords for every publication by every UCSF researcher in the system. We can use this to look at the most commonly used MeSH keywords across every researcher’s body of work. There are lots of caveats here (looking at all publications emphasizes past research interests over current ones; we’re not grouping related obscure MeSH terms with more popular ones; MeSH term assignment practices change over time; and this analysis ignores someone’s role as a first, middle, or last author). But this is certainly a start.
Here’s what I found, using the latest UCSF Profiles data:
Co-authorship networks can help us understand internal research collaboration patterns at UCSF. I used data from UCSF Profiles to create a visualization of (almost) every researcher currently at UCSF, and how their intra-UCSF co-authorship networks break out by department.
This visualization by department bears some more investigation than the previous one by school. Department of Medicine researchers are all over, collaborating with a wide variety of external departments. But an initial visual inspection suggests that almost all major departments have co-authorship relationships with members of other departments; some, like neurology, appear to form large standalone clusters, while others, like radiology, are more enmeshed in the work of others. This visualization flattens complex relationships into two dimensions, but it’s a starting point as we work to understand how UCSF collaborates.
View full-size visualization (PDF)
Co-authorship networks can help us understand internal research collaboration patterns at UCSF. I used data from UCSF Profiles to create a visualization of (almost) every researcher currently at UCSF, and how their intra-UCSF co-authorship networks break out by school.
Unsurprisingly, the School of Medicine takes up most of the space, and the visualization is probably most interesting in terms of what it might suggest about the smaller schools. Researchers from the Schools of Nursing and Dentistry form their own visible clusters, who often work with each other, but also have co-authorship relationships with researchers at the School of Medicine. But I was surprised by the School of Pharmacy, whose researchers form a main clusters in the bottom right, as well as additional clusters in the middle and top left, due to strong collaborative relationships with School of Medicine researchers.
View full-size visualization (PDF)
Some UCSF departments work more closely together than others. I looked at co-authorship patterns in papers published between January 2012 and November 2013, based on data in UCSF Profiles, and pulled out the UCSF departments that collaborate the most frequently. The results aren’t necessarily surprising. The Department of Medicine is huge, and their cross-departmental collaborations make up 8 of the top 10 collaborations, measured by volume. On the flip side, smaller groups with research areas similar to others make up many of the most common collaborations, by percentage; for example, one-third of papers by researchers primarily affiliated with the Proctor Foundation for Research in Opthalmology are co-authored with researchers from the Department of Opthalmology. I wouldn’t have necessarily guessed, however, connections like that between nursing and psychiatry.
Top UCSF cross-departmental collaborations, by volume
Top UCSF cross-departmental collaborations, by percentage
Details: Data is drawn from UCSF Profiles, and is based on a list of all publications listed on PubMed published between Jan 2012–Nov 2013, focusing on those whose authors include groups of researchers that have primary affiliations to more than one UCSF department. We counted only those publications from researchers with a listed department, and for the purposes of counting top cross-departmental collaborations by percentage, only those collaborations that generated 10 or more papers during the time period. No attempt was made to account for the widely varying sizes and scopes of different departments, the fact that researchers may have multiple departmental affiliations, or the fact that some publications may have been authored before the researchers were affiliated with their current primary departments at UCSF.
Photo: “Glasgow’s own superheroes having a smoke outside the Counting House” by Stephen Fyfe/Flickr, under CC-BY-NC-ND
When UCSF researchers collaborate between departments, how diverse are the collaborations? Here are the top 20 UCSF departments, ranked by the average numbers of UCSF departments their multi-departmental papers include as co-authors (from among the 39 departments whose researchers had a total of 25+ multi-departmental publications published between January 2012 and November 2013).
Details: Data is drawn from UCSF Profiles, and is based on a list of all publications listed on PubMed published between Jan 2012–Nov 2013 whose authors include groups of researchers with primary affiliations to more than one UCSF department. We counted only those publications from researchers with a listed department, and only those departments whose current associated researchers published 25+ publications in conjunction with current members of other UCSF departments between Jan 2012–Nov 2013. No attempt was made to account for the widely varying sizes and scopes of different departments, the fact that researchers may have multiple departmental affiliations, or the fact that some publications may have been authored before the researchers were affiliated with their current primary departments at UCSF. These are the top 20 departments, out of a total of 39 that match our criteria.
Some UCSF departments consistently reach out out to collaborate with other members of the UCSF community. Here are the top 20 UCSF departments whose researchers have the highest proportion of publications co-authored with members of other UCSF departments from among departments whose researchers had a total of 100+ publications published between January 2012 and November 2013.
Details: Data is drawn from UCSF Profiles, and is based on a list of all publications listed on PubMed published between Jan 2012–Nov 2013 whose authors include groups of researchers with primary affiliations to more than one UCSF department. We counted only publications from researchers with a listed department, and departments with 100+ publications by current associated researchers between Jan 2012–Nov 2013. No attempt was made to account for the widely varying sizes and scopes of different departments, the fact that researchers may have multiple departmental affiliations, or the fact that some publications may have been authored before the researchers were affiliated with their current primary departments at UCSF. These are the top 20 departments, out of a total of 42 that match our criteria.
Co-authorship networks give us a sense of the strength of research collaborations. We used co-authorship data to visualize how top HIV/AIDS research institutions worked with one another, based on publications from June 2012 to September 2013. UCSF collaborations are indicated via red lines.
Visualization details: Data includes all known publications related to HIV/AIDS between June 2012 and September 2013 that includes co-authors from two or more institutions. We map each author to their institution, and the size of each institution corresponds with the number of HIV/AIDS publications its members co-authored in that time; only the most prolific institutions are shown to ensure readability of the image. The width of the lines connecting institutions corresponds to the number of publications that include co-authors from both of these institutions. Collaborations with UCSF researchers are indicated with red lines. Colors indicate clusters of institutions that often publish collaboratively, based on network modularity.
UCSF researchers often work closely with one another, across departments. We used data from UCSF Profiles to visualize how different departments work together, based on co-authorship patterns.
Visualization details: Data is drawn from UCSF Profiles, and includes all publications co-authored by current UCSF researchers from two more departments and listed on PubMed. The size of each department corresponds with the number of publications that members have published that include partnerships with other departments. The width of the lines connecting departments corresponds to the number of publications between two departments. Colors indicate clusters of departments that often publish collaboratively, based on network modularity. No scaling is done to account for varying sizes of different departments.
BioMed 2.0 