Feeds:
Posts
Comments

Archive for the ‘Data availability’ Category

We’re pleased to present a guest post from data scientist Juan M. Banda, the lead author of an important, newly-available resource for drug safety research. Here, Juan shares some of the context behind the data descriptor in Scientific Data and associated data package in Dryad. – EH

_____

As I sit in a room full of over one hundred bio-hackers at the 2016 Biohackathon in Tsuruoka, Yamagata, Japan, the need to have publicly available and accessible data for research use is acutely evident. Organized by Japan’s National Biosciences Database Center (NBDC) and Databases Center for Life Science (DBLS), this yearly hackathon gathers people from organizations and universities all over the world, including the National Center for Biotechnology Information (NCBI) and the European Bioinformatics Institute (EBI), with the purpose of extending and interlinking resources like PubChem, PhenomeCentral, Bio2RDF, and PubAnnotation.

The end goal: finding better ways to access data that will allow researchers to focus on analysis of the data rather than preparation.

In the same spirit, our publication “A curated and standardized adverse drug event resource to accelerate drug safety research” (doi:10.1038/sdata.2016.26; data in Dryad at http://doi.org/10.5061/dryad.8q0s4) helps researchers in the drug safety domain with the standardization and curation of the freely available data from the Federal Food and Drug Administration (FDA) adverse events reporting system (FAERS).

FAERS collects information on adverse events and medication errors reported to the FDA, and is comprised of over 10 million records collected between 1969 to the present. As one of the most important resources for drug safety efforts, the FAERS database has been used in at least 750 publications as reported by PubMed and was probably manipulated, mapped and cleaned independently by the vast majority of the authors of said publications. This cleaning and mapping process takes a considerable amount of time — hours that could have been spent analyzing the data further.

Our publication hopes to eliminate this needless work and allow researchers to focus their efforts in developing methods to analyze this information.

OHDSIAs part of the Observational Health Sciences Initiative (OHDSI), whose mission is to “Improve health, by empowering a community to collaboratively generate the evidence that promotes better health decisions and better care,” we decided to tackle the task of cleaning and curating the FAERS database for our community, and the wider drug safety community. By providing a general common data model (CDM) and a general vocabulary to standardize how electronic patient data is stored, OHDSI allows its participants to join a research network with over 655 million patients.

With a significant fraction of the community’s research being focused on drug safety, it was a natural decision to standardize the FAERS database with the OMOP vocabulary, to allow all researchers on our network access to FAERS. Since the OMOP vocabulary incorporates general vocabularies such as SNOMED, MeSH, and RxNORM, among others, the usability of this resource is not limited to participants of this community.

In order to curate this dataset, we took the source FAERS data in CSV format and de-duplicated case reports. We then performed value imputation for certain fields that were missing. Drug names were standardized to RxNorm ingredients and standard clinical names (for multi-ingredient drugs). This mapping is tricky because some drug names have spelling errors, and some are non-prescription drugs, or international brand names. We achieved coverage of 93% of the drug names, which in turn cover 95% of the case reports in FARES.

For the first time, the indication and reactions have been mapped to SNOMED-CT from their original MedRA format. Coverage for indications and reactions is around 64% and 80%, respectively. The OMOP vocabulary allows RxNorm drug codes as well as SNOMED-CT codes to reside in the same unified vocabulary space, simplifying use of this resource. We also provide the complete source code we developed in order to allow researchers to refresh the dataset with the new quarterly FAERS data releases and improve the mappings if needed. We encourage users to contribute the results of their efforts back to the OHDSI community.

With a firm commitment to making open data easier to use, this resource allows researchers to utilize a professionally curated (and refreshable) version of the FAERS data, enabling them to focus on improving drug safety analyses and finding more potentially harmful drugs, as a part of OHDSI’s core mission.

OHDSI_still2

Still from OHMSDI video

The data:

http://doi.org/10.5061/dryad.8q0s4

A full description of the dataset in Scientific Data:

http://www.nature.com/articles/sdata201626

 

— Juan M. Banda

Read Full Post »

On May 24, we held the first virtual Dryad Community Meeting, which allowed us to connect both with our membership and with the larger open data community, far and wide. The theme was “Leadership in data publishing: Dryad and learned societies.”

Following an introduction and update about Dryad from yours truly, we heard about the experiences from representatives of three of Dryad’s member societies.

All three societies require that data be archived in an appropriate repository as a condition of publication in their journals. Yet, they have each taken considerable time and effort to develop policies that address the needs and concerns of their different communities.

Bruna spoke about working with an audience that routinely gathers data for very long-term studies. For many Biotropica authors, embargoes are seen as an important prerequisite for data publishing. Their data policy “includes a generous embargo period of up to three years to ensure authors have ample time to publish multiple papers from more complex or long-term data sets”. Biotropica’s policy also recommends those “who re-use archived data sets to include as fully engaged collaborators the scientists who originally collected them”. To address initial resistance to data archiving, and to build understanding and consensus, Biotropica “enlisted its critics” to contribute to a paper discussing the pros and cons of data publication. Out of this process emerged an innovative policy that went into effect at the start of 2016.

Meaden, by contrast, noted that only 8% of Proceedings B authors elect to embargo data in Dryad, and the standard embargo is for only one year after publication. She credited clearer author instructions and a data availability statement in the manuscript submission system as key elements that have increased the availability of data associated with Royal Society publications.

Newton discussed BES’ move from “encouraging data publication” in 2012 to requiring it in 2014. As shown below, this resulted in an impressive increase in the availability of data. Next, the society is looking to develop guidance on data reuse etiquette. Newton noted that this effort would “need to be community-led.”

BES_data_preservation

Slide from Erika Newton’s presentation, illustrating the rise in data deposits for BES journals as associated with changing data policy.

While each speaker reported on unique challenges, all shared commonalities, such as:

  • involving the specific community in policy decisions
  • incrementally increasing efforts to make data available
  • the importance of clear author instructions 

We greatly appreciate the excellent contributions from the panelists, as well all the members and other attendees who participated and contributed to the lively Q&A.

We are also pleased that the virtual format was well received. In our follow-up survey, many of the attendees said they found it easy to ask questions and appreciated the ability to join remotely.

Our aim is that these meetings continue to be a valued forum for our diverse community of stakeholders to share knowledge and discuss emerging issues. If you have suggestions on topics for future meetings, or an interest in becoming a member, please reach out to me at director@datadryad.org.

dryad_members

 

Read Full Post »

The following is a guest post from science journalist John Bohannon. We asked him to give us some background on his recent dataset in Dryad and the analysis of that data in Science. What stories will you find in the data? – EH

_______

Scihub_raven

Sci-Hub is the world’s largest repository of pirated journal articles. We will probably look back and see it as inevitable. Soon after it became possible for people to share copyrighted music and movies on a massive scale, technologies like Napster and BitTorrent arrived to make the sharing as close to frictionless as possible. That hasn’t made the media industry collapse, as many people predicted, but it certainly brought transformation.

Unlike the media industry, journal publishers do not share their profits with the authors. So where will Sci-Hub push them? Will it be a platform like iTunes, with journals selling research papers for $0.99 each? Or will Sci-Hub finally propel the industry into the arms of the Open Access movement? Will nonprofit scientific societies and university publishers go extinct along the way, leaving just a few giant, for-profit corporations as the caretakers of scientific knowledge?

There are as many theories and predictions about the impact of Sci-Hub as there are commentators on the Internet. What is lacking is basic information about the site. Who is downloading all these Sci-Hub papers? Where in the world are they? What are they reading?

48 hours of Sci-Hub downloads. Each event is color-coded by the local time: orange for working hours (8am-6pm) and blue for the night owls working outside those hours.

Sometimes all you need to do is ask. So I reached out directly to Alexandra Elbakyan, who created Sci-Hub in 2011 as a 22 year-old neuroscience graduate student in Kazakhstan and has run it ever since. For someone denounced as a criminal by powerful corporations and scholarly societies, she was quite open and collaborative. I explained my goal: To let the world see how Sci-Hub is being used, mapping the global distribution of its users at the highest resolution possible while protecting their privacy. She agreed, not realizing how much data-wrangling it would ultimately take us.

Two months later, Science and Dryad are publicly releasing a data set of 28 million download request records from 1 September 2015 through 29 February 2016, timestamped down to the second. Each includes the DOI of the paper, allowing as rich a bibliographic exploration as you have CPU cycles to burn. The 3 million IP addresses have been converted into arbitrary codes. Elbakyan converted the IP addresses into geolocations using a database I purchased from the company Maxmind. She then clustered each geolocation to the coordinates of the nearest city using the Google Maps API. Sci-Hub users cluster to 24,000 unique locations.

The big take-home? Sci-Hub is everywhere. Most papers are being downloaded from the developing world: The top 3 countries are India, China, and Iran. But the rich industrialized countries use Sci-Hub, too. A quarter of the downloads came from OECD nations, and some of the most intense download hotspots correspond to the campuses of universities in the US and Europe, which supposedly have the most comprehensive journal access.

But these data have many more stories to tell. How do the reading habits of researchers differ by city? What are the hottest research topics in Indonesia, Italy, Brazil? Do the research topics shift when the Sci-Hub night owls take over? My analysis indicates a bimodal distribution over the course of the day, with most locations surging around lunchtime, and the rest peaking at 1am local time. The animated map above shows just 2 days of the data.

Something everyone would like to know: What proportion of downloaded articles are actually unavailable from nearby university libraries? Put another way: What is the size of the knowledge gap that Sci-Hub is bridging?

Download the data yourself and let the world know what you find.

The data:

http://dx.doi.org/10.5061/dryad.q447c

My analysis of the data in Science:

http://www.sciencemag.org/news/2016/04/whos-downloading-pirated-papers-everyone

 

 — John Bohannon

Read Full Post »

2015While gearing up for the Dryad member meeting (to be held virtually on 24 May – save the date!) and publication of our annual report, we’re taking a look at last year’s numbers.

2015 was a “big” year for Dryad in many respects. We added staff, and integrated several new journals and publishing partners. But perhaps most notably, the Dryad repository itself is growing very rapidly. We published 3,926 data packages this past year — a 44% increase over 2014 — and blew past the 10,000 mark for total data packages in the repository.

Data package size

Perhaps the “biggest” Dryad story from last year is the increase in the mean size of data packages published. In 2014, that figure was 212MB. In 2015, it more than doubled to 481MB, an increase of a whopping 127%.

This striking statistic is part of the reason we opted at the beginning of 2016 to double the maximum package size before overage fees kick in (to 20GB), and simplified and reduced our overage fees. We want researchers to continue to archive more (and larger) data files, and to do so sustainably. Meanwhile, we do continue to welcome many submissions on the smaller end of the scale.

boxplot_logscale_labels

Distribution of Dryad data package size by year. Boxplot shows median, 1st and 3rd quartiles, and 95% confidence interval of median. Note the log scale of the y-axis.

In 2015, the mean number of files in a data package was about 3.4, with 104 as the largest number of files in any data package. To see how times have changed, compare this to a post from 2011 (celebrating our 1,000th submission), where we noted:

Interestingly, most of the deposits are relatively small in size. Counting all files in a data package together, almost 80% of data packages are less than one megabyte. Furthermore, the majority of data packages contain only one data file and the mean is a little less than two and a half. As one might expect, many of the files are spreadsheets or in tabular text format. Thus, the files are rich in information but not so difficult to transfer or store.

We have yet to do a full analysis of file formats deposited in 2015, but we see among the largest files many images and videos, as would be expected, but also a notable increase in the diversity of DNA sequencing-related file formats.

So not only are there now more and bigger files in Dryad, there’s also greater complexity and variety. We think this shows that more people are learning about the benefits of archiving and reusing multiple file types, and that researchers (and publishers) are broadening their view of what qualifies as “data.”

Download counts

2015speciesSo who had the biggest download numbers in 2015? Interestingly, nearly all of last year’s most-downloaded data packages are from genetics/genomics. 3 of the top 5 are studies of specific wild populations and how they adapt to changing circumstances — Sailfin Mollies (fish), blue tits (birds), and bighorn sheep, specifically.

Another top package presents a model for dealing with an epidemic that had a deadly impact on humans in 2015. And rounding out the top 5 is an open source framework for reconstructing the relationships that unite all lineages — a “tree of life.”

In 5th place, with 367 downloads:

In 4th place, with 601 downloads:

In 3rd place, with 1,324 downloads:

In 2nd place, with 1,868 downloads:

And this year’s WINNER, with 2,678 downloads:

The above numbers are presented with the usual caveats about bots, which we aim to filter out, but cannot do with perfect accuracy. (Look for a blog post on this topic in the near future).

As always, we owe a huge debt to our submitters, partners, members and users for supporting Dryad and open data in 2015!

Read Full Post »

We are delighted to announce the launch of a new partnership with The Company of Biologists to support their authors in making the data underlying their research available to the community.

COBNewLogo300dpiThe Company of Biologists is a not-for-profit publishing organization dedicated to supporting and inspiring the biological community. The Company publishes five specialist peer-reviewed journals:

The Company of Biologists offers further support to the biological community by facilitating scientific meetings, providing travel grants for researchers and supporting research societies.

Manuscript submission for all COB journals is now integrated with data submission to Dryad, meaning COB authors can conveniently submit their data packages and manuscripts at the same time. Dryad then makes the data securely available to journal reviewers, and releases them to the public if/when the paper is published.

We congratulate The Company of Biologists on taking this important step to help facilitate open data. To learn more about how your organization or journal can partner with Dryad, please contact us.

Read Full Post »

Our latest featured data package is from Alexandra Swanson and colleagues at the Snapshot Serengeti project, and accompanies their peer-reviewed article in Scientific Data.  It provides a unique resource for studying one of the world’s most extraordinary mammal assemblages and also for studies of computer vision and machine learning. In addition, data from Snapshot Serengeti is already being used in biology and computer science classrooms to enable students to work on solving real problems with authentic research data.

 lion

Snapshot Serengeti, CC BY-NC-SA 3.0

The raw data (which are being made available from the University of Minnesota Supercomputing Institute) consist of 1.2 million sets of images collected between February 2011 and May 2013 from 225 heat and motion triggered cameras, operating day and night, distributed over 1,135 sq. km. in Serengeti National Park in Tanzania.  This staggering trove of images was classified by 28,040 registered and ~40,000 unregistered volunteers on Snapshot Serengeti (a Zooniverse project) according to the species present (if any), the number of individuals, the presence of young, and what behaviors were being displayed, such as standing, resting, moving, eating, or interacting.

Remarkably, this vast army of citizen scientists was classifying the images faster than they were being produced, and each image set was classified on average by nine different volunteers.  This led to consensus classifications with high accuracy, 96.6% for species identifications relative to an expert-classified gold set.  Of the more than 300,000 image sets that contain animals, 48 different species were seen, including rare mammals such as the aardwolf and the zorilla.

zorilla

zorilla (image from Snapshot Serengeti CC BY-NC-SA 3.0)

The Dryad data package includes the classifications from all the individual volunteers, the consensus classifications, information about when each camera was operational, and the expert classification of 4,149 image sets as a gold standard.

References:

  • Swanson et al. (2015) Snapshot Serengeti, high frequency annotated camera trap images of 40 mammalian species in an African savannah. Scientific Data.  http://dx.doi.org/10.1038/sdata.2015.26
  • Swanson et al. (2015) Data from: Snapshot Serengeti, high frequency annotated camera trap images of 40 mammalian species in an African savannah. Dryad Digital Repository http://doi.org/10.5061/dryad.5pt92

Read Full Post »

The reason why Dryad is in the business of archiving, preserving, and providing access to research data is so that it will be reused, whether for deeper reading of the publication, for post-publication review, for education, or for future research. While it’s not yet as easy as we would like to track data reuse, one metric that is straightforward to collect is the number of times a dataset has been downloaded, and this is one of two data reuse statistics reported by our friends at ImpactStory and Plum Analytics.

2014 with fireworks

The numbers are very encouraging. There are already over a quarter million downloads for the 8,897 data files released in 2014 (from 2,714 data packages). That’s over 28 downloads per data file. While there is always the caveat that some downloads may be due to activity from newly emerged bots that we have yet to recognize and filter out, we think it is safe to say that most of these downloads are from people.

To celebrate, we would like to pay special tribute to the top five data packages from 2014, as measured by the maximum number of downloads for any single file (since many data packages have more than one) at the time of writing. They cover a diversity of topics from livestock farming in the Paleolithic to phylogenetic relationships among insects. That said, we are struck by the impressively strong showing for plant science — 3 of the top 5 data packages.

In 5th place, with 453 downloads

In 4th place, with 581 downloads

In 3rd place, with 626 downloads

In 2nd place, with 4,672 downloads

And in 1st place, with a staggering 34,879 downloads

Remarkably, given the number of downloads, this last data package was only released in November.

We’d like to thank all of our users, whether you contribute data or reuse it (or both), for helping make science just a little more transparent, efficient, and robust this past year. And we are looking forward to finding out some more of what you did with all those downloads in 2015!

 

 

 

 

Read Full Post »

Older Posts »

Follow

Get every new post delivered to your Inbox.

Join 10,502 other followers