How do you find science papers? Tips for connecting to a new field.

In my last blog post, I nerded out over some useful options that PubMed can provide to the well-informed user looking for new science literature. Now that you’ve become a master of finding papers when you already know what you’re looking for, you may next need to know how to find entirely new pockets of literature and become quickly oriented within them. This is a slightly different type of task, and you may want to try an alternative approach. The new resource I want to tell you about is called Connected Papers, and I am excited about the way it combines functionality with a neat visualization approach.

The purpose of Connected Papers is to take a publication you suggest and build a virtual node map of papers that may be connected to it by similarity so as to give you a collection of papers contemporary to the one you searched. Similar papers may include some that are relevant to your work but may have lower visibility because they have not been cited as much and would otherwise be difficult to find. The force diagram built from your query allows you to explore these connected papers and uses colors and line thickness to give these connections more nuance. Papers (circles) are weighed by citation (bigger circle = more citations) and recency (darker color = more recent) and are clustered and joined by similarity, creating a visual map of related research for you to explore.

“The purpose of Connected Papers is to take a publication you suggest and build a virtual node map of papers that may be connected to it by similarity so as to give you a collection of papers contemporary to the one you searched.”

-- Anny Gano, postdoctoral fellow

Particularly cool is the way Connected Papers builds the node diagram. It looks for overlapping bibliographies and co-citations to create clusters of more or less relevant literature as opposed to just being a straight citation tree. This means you do not need to rely on direct links between papers to discover relevant literature. You will instead get a more thorough overview of the field, with papers from multiple research groups available for your perusal. 

As an example, my field of expertise is neuroscience. Let’s say I recently read a really fun paper about some newly proposed neural mechanisms of lycanthropy (Fun fact time: Lycanthropy is an actual clinical disorder, but for the sake of this imaginary example, let’s talk about the more common folklore meaning of the word: turning into a werewolf!) Having read this werewolf paper, I thought this was a really cool topic that I might want to pursue in the next grant cycle. My next step likely would be to google the authors and their laboratory, read some of their recent work and then to read some of the top papers they cited, all of which would likely be from the animal transformation field with a focus on the wolf. 

Here’s the cool thing about Connected Papers – if I plugged my source lycanthropy paper into its search bar, I would not only see the relevant werewolf research, but I would also see papers that cite the same animal transformation literature in general. This feature could lead me to some interesting new reading that I might have otherwise ignored – for example, a seminal paper describing a protein that can trigger a human vampire’s transformation into a bat! If not for checking out these similar papers, I might never have seen this paper because the animal model they use (bat) is so different than the lupine work I was reading before. Finding this paper could be the inspiration I need to come up with my innovative grant idea: the use of the novel bat protein as a therapeutic tool to make werewolf transformation voluntary, freeing them from being under the control of the full moon!

The somewhat colorful example aside, I actually recently used Connected Papers when trying to orient myself in a new lab project, and I definitely intend to use it next time I write a review paper or immerse myself in a novel area. It should be noted that in order to get the best use out of this tool, you should start with a specific paper that sparks your interest. Using the search as a keyword finder is not the most optimal use of Connected Papers – the search focuses specifically on paper titles, so your results may not be precise if you are trying to use it to search a general concept.

The designers Alexander Tarnavsky Eitan, Eddie Smolyansky and Itay Knaan Harpaz are all experts from different technical fields and started this as a creative side project among friends, which is the kind of energy that frequently creates great products that fill a niche. They have already swiftly made improvements to the website based on user feedback, such as adding a highlighted similarity path when you hover a pointer over a particular paper in the diagram that connects it to your original source paper. I look forward to seeing what other tools they develop!

# # # 

A big thank you to Dr. Melissa Conti Mazza (NIH) for giving me the tip to check out this tool, and to Alexander Tarnavsky Eitan of Connected Papers for graciously giving me his time and corrections for this piece.