Expanding Meta’s knowledge graph to build a researcher’s guide to brain organoids
By Burcin Ikiz, Ph.D. & Michael Czerwinski, Ph.D.
One of the biggest challenges for a researcher is entering a new field. Whether one is a graduate student starting their thesis work, an assistant professor building their lab, or a seasoned researcher writing a grant, it can be overwhelming — and time-consuming — to dive deep into the literature and find the right research papers, reviews, and protocols to get up to speed with an unfamiliar topic.
Over the last five months, we developed a prototype to address this challenge by generating a map of biomedical research. The map, or knowledge graph, goes beyond a simple catalog of articles to include methods, data sets, compounds, and other research outputs that are mentioned in papers. …
The Meta team reflects on their best work of 2020 and shares what’s coming in 2021
In the past year, Meta has given users more ways to get to the research that counts faster than ever before. With new developments like more precise feed design, personalized ranking, and the ongoing addition of new content, Meta is increasingly becoming a place where biomedical researchers can get a clearer view of rapidly changing science beyond published articles and preprints, no matter their area of interest.
While Meta may be known for being powered by machine learning, there is also a whole team of people behind the scenes, dedicated to bringing the best data and best experiences to Meta users. …
How a Meta feed is constructed with pertinent papers to keep researchers up to date
Meta’s personalized feeds are at the heart of research discovery, providing researchers with a faster way to organize, understand, and explore science. They operate differently from an RSS feed or email list, and instead draw on a suite of AI-based technologies to deliver the latest papers and preprints.
What is a feed?
A Meta feed is a collection of research articles and preprints (and soon, other content types like datasets, protocols, software, and clinical trials), presented as a timeline that updates automatically as new content becomes available. Feeds aggregate scientific information from many sources in real time — papers from PubMed, and preprints from bioRxiv, medRxiv, q-bio, and other preprint servers as they are released by different databases and repositories — making it fast and easy to follow biomedical research topics and the information associated with them at every stage of the discovery process. …
Meta Data Story
We surveyed the literature to discover the fastest growing areas of disease research , and how COVID-19 changed the landscape.
To gauge how biomedical research has evolved in recent years, and especially in light of the current pandemic, Meta analysts examined two questions:
- Between 2015 and 2020, which areas of disease research have grown the fastest in terms of paper production?
- How has COVID-19 affected this progression?
The analysis is based on publication data in Meta. Meta’s store of published biomedical papers and preprints — including each record’s related metadata — surpasses 30 million, with some 4,000 additions each day. This dataset provides a detailed view of research activity, output, and publishing patterns.
The COVID-19 Effect: Infections & Respiratory Tract Diseases Research Not Only Grew, They Dominated
We categorized papers using disease designations from the Medical Subject Headings, or MeSH, used by the National Library of Medicine for indexing, cataloguing, and searching biomedical literature. Then we compared the rate of growth in paper and preprint volume over a five-year window to capture longer-term publishing trends. The five disease fields that experienced the highest growth are plotted in Figure 1. …
COVID-19
Potent mini antibodies could lead the way to faster, cheaper antiviral drugs
Moving quickly to get COVID-19 therapeutics into the clinic, the U.S. Food and Drug Administration (FDA) recently granted emergency approval to two antibody treatments for COVID-19, developed by Eli Lilly and Regeneron. The drugs generally contain one or more disease-fighting proteins called monoclonal antibodies that mimic those made by the human immune system. So far, they are proving safe, making researchers even more optimistic about the potential of lab-made antibodies to prevent and treat SARS-CoV-2 infections.
A recent article in Nature Biotechnology pegged the number of antibody treatments for COVID-19 that are undergoing human testing at 13, with at least another 150 in the discovery phase. The Coronavirus Immunotherapeutic Consortium (CoVIC), a U.S.-based nonprofit organization providing independent evaluation of antibody therapeutics, has a slightly higher tally. It has received more than 180 therapeutic antibody candidates for evaluation from academics, nonprofits, biotech and pharmaceutical companies around the world. …
WEBINAR
Join our free webinar on November 4th to learn how baby turtles and narcoleptic dogs led to careers in science
At the best of times, a career transition can be difficult, especially when you’re jumping to a whole new field. Meta’s product application scientists, Burcin Ikiz and Michael Czerwinski, can shed some light on what that process can look like.
Because of Michael and Burcin’s collective experience in academia and industry, their job is to wear their scientist hats while working with our teams to develop Meta. In their current roles, they get to problem solve every day and help fill a major need in the biomedical research community. …
COVID-19
Scientists are scouring SARS-CoV-2 genomes for clues to the virus’s origins, spread and evolution. What’s written into its RNA?
PRODUCT NEWS
How we used BioBERT to surface the papers you most want to see
At Meta, we are helping biomedical researchers keep up with the latest scientific preprints and articles through feeds powered by machine learning models. Recently, we introduced Matched to You feed sorting, in which articles are now organized by their relevance to you based on your interaction with the content in a feed.
The Meta Personalized Ranking Algorithm
Our personalized ranking algorithm is a content-based recommender system model that predicts what scientific articles users would be most interested in.
Let’s use a feed on COVID-19 as an example. Below is the composition for a feed named “COVID-19” based on a number of keywords and concepts. This query will retrieve papers that are either tagged with ‘COVID 19’, or with the concept 2019-NCoV, or with the concept COVID-19, and so on. …
PRODUCT NEWS
Watch highlights from our webinar on how to save time keeping up with new research
Unleashing the power of Meta was the focus of a recent webinar, and the first in a series. In roughly 30 minutes, Product Manager Erin Hoops provided an introduction to Meta, answered questions, and gave an overview of the exciting changes and new features we have planned for Meta in the near future.
By watching this recording, you can easily get started with using Meta. The video is divided into chapters to help you readily access the part of Meta you want to learn about (on the play bar, the dots indicate chapters, and the chapter playlist will pop up when you click the second icon from the right). …
Go deeper into the transformative research honored by one of science’s biggest prizes.
Last week, the 2021 Breakthrough Prizes were awarded to biomedical researchers David Baker, Catherine Dulac, Dennis Lo, and Richard Youle. Their discoveries — on the design of synthetic proteins, the neural circuits of parenting, prenatal testing for genetic disease, and the role of key proteins in hereditary Parkinson’s — have been transformative, overturning longstanding ideas in biology and opening up promising new avenues of exploration.
Join us in celebrating these four remarkable scientists, and use Meta’s Breakthrough Prize-related feeds to go deeper into the research areas the prizewinners have shaped and follow their science as it evolves.
David Baker, Protein Designer
David Baker, PhD, designs and builds proteins from scratch, a feat of scientific derring-do made possible by decades of research into understanding how chains of amino acids twist, bend, and fold into their final shape. …
About
