A synthesis journal club

We thought we would share how our lab graph has allowed us to conduct a new type of journal club that we made up, called a "synthesis journal club."

Usually, journal clubs focus on a single research article. What do the authors want us to know about their findings? Our lab graph lets us take a different tack: our journal clubs are now organized around a question rather than around a single article.

We start by listing the research questions that have come up in our notes recently. In our graph, we do this querying for Question pages in a particular time window.

 

We discussed this during the "lab updates" section of our weekly group meeting.

 

59 question pages from the past four months!

From this list, everyone selects their favorite questions, to make a shortlist. Two of us then classified the shortlist questions roughly based on the following criteria:

• Is the question of general interest to the lab?

• Is the answer already established? (If so, then we should just google the answer)

• Do we think there are published articles that inform this question? (If not, maybe we should start our own research project in the area!)

 

In the end, our triage left us with one general-interest question, the one at the bottom.

Once we settled on a question, we each listed all the research articles that we thought might inform the question. (E.g. searching "integrin" in the lab graph and seeing what pops up.)

 

Each of these articles is linked to our lab Zotero collection, thanks to an open-source plugin.

Each person claimed one or two articles. Importantly, rather than reading the article front to back, we each scanned just for the figure(s) that pertained to our research question, and focused on those.

 

For example, I took notes on this article in my daily notes. I converted the most salient result into an evidence page.

 

The evidence page contains the model system, approach, and observation in past tense, plus contextual information.

We each came to journal club with a different piece of the puzzle. It was very exciting anticipating that you only had a piece of the larger picture!

We can then query to collect all the new "discourse pages" created in the question page,

 

Query: find all the discourse pages that were referenced inside our Question page.

Then! We send these pages to a blank canvas:

 

"Send pages to a Canvas"

 

All of our discourse pages, waiting to be arranged into a coherent picture!

And now the journal club actually starts 😁

I guess this makes it a "flipped" journal club?

During the journal club, we start with a short intro about the motivation for the question.

 

Setting up the question

Then each person describes their evidence page(s) and puts them in context by describing the model system and the method. Importantly, the authors' motivation for writing the paper doesn't come up, because we're slicing one observation from their article for our synthesis.

 

Different lab members take notes while one person is talking about their evidence pages. Tellingly, we referenced existing discourse pages from previous journal clubs to reinforce our point.

When necessary, the person draws on the canvas to illustrate their point, and uses the evidence page to ground their point.

 

Here we are trying to work through whether two findings are consistent or contradictory.

At the end, we try to put the pieces together, with a drawing, a couple of sentences, and a couple of "synthesis-level" claims based on the accumulated evidence.

 

Notably, our focus shifted beyond the original question (which, it turns out, had a well-established answer) to a more complex, related question.

 

Those hyperlinks point to our claim pages, or evidence pages that substantiate the claim.

And finally! I used our really experimental GPT4 integration to ask AI to summarize our findings.

 

Brand new interface! Very experimental

 

About 80% of GPT4's evidence statements seemed reasonable 😄 Which means our collective brainpower is still necessary for the sensemaking to happen!

I'm very excited that this synthesis journal club has shown itself to be:

• Targeted: We filter articles for how they address our chosen research question.

• Reusable: The evidence and claim pages become persistent objects that we return to, reference, and add to.

• Scalable: Such a journal club format should work for much larger groups of people interested in a common research question.

This type of journal club should help new researchers get up to speed on the state of a research question much faster!