How to use scite to better understand any topic in the world
In 2011 Marc Andreesen wrote that “software is eating the world.” The foresight of Andreesen should not be surprising given his involvement in launching the first widely used web browser Netscape. Yet still, somehow we are surprised when software takes over another part of our lives, such as the refrigerator, speakers, or even doorknobs. But there is one place in the world that software has had a surprisingly hard time digesting: scientific publishing.
Scientific publications are some of the world’s most important pieces of work. They describe life-saving vaccines, data and evidence about the origin of life, technologies used by billions of people around the world, and basically anything you can think of. There are even scientific publications looking at dog farts and how to properly kick a football. Despite their immense importance and utility to nearly all aspects of human life, most scientific publications have for the most part sat untouched.
There are a lot of reasons for this but it boils down to two main reasons: 1) scientific articles are under restrictive copyright and not easily “edible” and 2) scientific papers are in formats that make it hard for machines to eat. It’s almost as if machines have gluten intolerance and scientific papers are 90% gluten. To put it mildly, this is unfortunate as there is a wealth of information contained within scientific papers that could make the world a better place.
We built scite in an effort to unlock this wealth of knowledge by unleashing an army of machines on scientific publications. To process millions of articles and identify which ones support each other, which ones dispute each other, or more generally how papers interact with one another.
This has resulted in an entirely novel way to look at nearly any topic of interest and become easily acquainted with the literature at nearly the same capability as an expert in the field. To demonstrate its usefulness, I recently asked on LinkedIn for anyone to post something they were interested in and I would share with them a custom data dashboard showing which articles related to that topic were most supported, most disputed, or overall cited. This resulted in fascinating searches like “Cell death mechanisms during electrosurgical ablation,” to which I returned within a few minutes a custom dashboard identifying the most supported and disputed article in the field. Another person suggested, “Aducanumab.” Again, in a few minutes, I could identify all relevant articles mentioning Aducanumab and easily sort which had received the most supporting or disputing evidence by subsequent studies in the literature. People responded, “This is very useful” “Wow, this is super impressive!” and my favorite, “I now understand scite a lot better.”
Given that people on LinkedIn can be quite friendly, I repeated this ask on Twitter as a control, and immediately the requests started pouring in. In exchange for creating these dashboards, I explicitly asked people to share if it captured their field well, they learned something new, or if something was unclear. One statistics Ph.D. student replied, “Pretty solid,” which if you know any statisticians is high praise. Another replied, “Incredible! Super useful. I imagine this could be added on to systematic review software like Covidence or Distiller during abstract screening… would be a great function.”systematic review software like Covidence or Distiller during abstract screening…
The funny thing is that this magic is not something that I alone can do, it’s something anyone can do using scite. Here I walk you through how to create a custom dashboard on scite so that you can supercharge your research without having to get a Ph.D. in each topic.
Step 1: Identify the relevant articles
There are currently four ways you can make a custom dashboard:
- Importing your Zotero library
- Importing your Mendeley library
- Manually entering a comma-separated list of valid DOIs
- Uploading a CSV file containing a comma-separated list of valid DOIs
After you make your selection, simply click “Submit” and the dashboard will be created in a matter of seconds.
Got it. So, what am I looking at here…?
These custom dashboards communicate aggregated information about the group of DOIs you selected. In the top section, you can see:
- Metadata about your dashboard (name and description)
- The notification Set Alert bell next to the title allows you to receive email notifications if we detect new citations to any paper in this set.
- A count of the number of publications that comprise this dashboard
- The total supporting, mentioning, and disputing citations these articles have received
- A count of the number of editorial notices these publications have received
- The 2-year, 5-year, and all-time scite Index for this group of publications
Hold on, what exactly is the scite Index?
The scite Index (SI) measures how supported a group of publications are, and is calculated using the following formula:
The formula for the scite Index.
For example, the 2019 2-year SI includes citations to articles published in 2018 and 2019
In the defined window, there must be at least 100 supporting and/or disputing cites to receive a scite Index score.
Moving on, if we look further below on the dashboard, we find that we can visualize the trends in the scite Index over time for this group of publications.
Further below, we see a table of all the articles in this dashboard, along with the respective citation tallies they have received from other publications.
The table allows you to see which articles in that group are most disputed, supported, and so on. You can also filter them by their title to find specific ones. Clicking on the title of the article takes you to its corresponding scite report.
Got it… Do you have a concrete example?
Our custom dashboards are a powerful way to track groups of articles. Let’s walk through an example where we create one from our search results.
Tracking articles on COVID-19 published in Nature
As we mentioned previously, one of the most powerful workflows is being able to create a custom dashboard from search results.
Here’s how it looks at a high level:
- Search for something in scite (or PubMed)
- Export search results to CSV
- Copy the list of DOIs in the exported result set
- Use them to create the custom dashboard
Let’s walk through an example from within scite itself.
- Search for relevant publications
Suppose we were interested in COVID-19 related articles published in the journal Nature. We can perform such a search within scite:
2. Export search results
As you see in the screenshot, you can select Export CSV and generate a CSV of up to 1,000 results. If we download the CSV, we see a number of results and a column called DOI.
3. Copy list of DOIs from exported results
From the CSV, we can select the DOIs and copy them to our clipboard.
4. Create Custom Dashboard
Now, with those DOIs copied to our clipboard, we can return to our form for creating a custom dashboard, and manually paste those DOIs into the field.
Finally, View Dashboard
And at this point, we can submit the form and see that our dashboard is created. You can access this one here: https://scite.ai/dashboard/nature-covid-19-search-export-9Q2
Next, you can go beyond just the aggregate metrics about these papers and leverage the table at the bottom to drill down on them by their citation counts, titles, and quickly access their reports.
You can also configure an email alertby clicking on the bell at the top to receive email alerts when we detect new citations for any paper in this set.
