Crowdsourcing and the value of curation – Mendeley pairs with F1000 Prime to recommend great research.

Image via psd

What makes Mendeley more than just a reference manager is the community of researchers who use our tool to share research, recommend papers to others, and collaboratively work together. In practical terms, what this means is that the work of finding and organizing a collection of papers about a specific topic can be shared by a group of people via Mendeley Groups. Hundreds of thousands of you have created groups and saved your colleagues hundreds of thousands of hours, which I like to think has made research progress at least a little faster. But the benefits of groups go beyond crowdsourced literature discovery – you can annotate and comment on the collection of papers as well. There’s recently been quite a lot of activity around the idea of commenting on research, given the high-profile launch of Pubmed Commons and the growing attention given to tools such as PeerLibrary and PubPeer. We believe that post-publication commenting on research can become an important part of scholarly communication, but the past few years of our experience and the experience of PLOS shows that much of the literature doesn’t elicit comments, for whatever reason. Here’s where we feel a combination of traditional editorial insight into what is attention getting can be blended with the crowdsourced approach to yield high quality comments on high quality research. Our first experiment with this is the launch of F1000 Prime recommendations on Mendeley. Check out these great papers chosen by experts in Bioinformatics and Computational Biology in the first of what I hope will be a series of F1000 Prime groups. You can read F1000’s take here.

This content normally requires a subscription, but is made available free in the group. Just follow the group to get updates!

Fancy some Pi? Hack In!

Mendeley Hack Day 2013

Hack Days have a strong tradition here at Mendeley, and every month the team has a day to pursue whatever projects they like, whether it’s a product improvement, a cool idea they have for Mendeley, the chance to learn some new skills, or do something fun!

This month was no different, except that we decided to have a Raspberry Pi themed day, so the Mendeleyans could play around with the credit-card-size little computer that’s been so talked about. Even those without a programming bone in their bodies, such as yours truly, got in on the action, even if it was just by baking a giant raspberry pie/cake/scary calorie moutain…


Here are some of the cool projects we came up with, the first of many to come I’m sure! You can also check out some photos of the day on our Flickr page and a video on our Youtube page, and we’d love to hear if you have any ideas for cool hacks we could try next time, whether fruit-based or not.

Live Community Feed

João from the Community team got together with Carles Pina and Chris Barr to hook up a live feed to a large TV screen here at Mendeley HQ. To start with they wrote scripts and designed a user interface so that we could show live Tweets mentioning Mendeley, plus the feedback that we get from users applying to become advisors. We also have a great map showing user activity around the world, and plan to continue adding things like Facebook and YouTube feeds, plus many more so we can be even more in touch with the Mendeley community.



Victor Fernandez from our development team is a big fan of vintage games, and so he decided to learn his way around the Raspberry Pi by seeing how well it would cope with emulating various platforms, from DOS to Apple2. He installed a custom OS version of the Raspberry Pi which has an interface that lets you plug a controller directly in without need for a keyboard or mouse, and we were treated to some classic demos including 1980s gems such as Street Fighter, Ultima, Cave Story, and Batman.


Lego Mindstorms EV3 Robot

Adding Lego to a project usually means multiplying the fun factor, and when you end up with a cool-looking robot at the end of your Hack Day it’s hard to argue with that. George Kartvelishvili teamed up with honorary Mendeleyan and Lego nutcase Tom Atkinson to build the cool Mindstorms Ev3 robot and hook it up to the Raspberry Pi so that it could be controlled simultaneously from either a laptop or a Bluetooth app. Next time we’re on the lookout for ideas of how we could incorporate Mendeley data and themes into the robotic fun…


Two shades of grey

And not content with his robotic enterprises, George also found time to connect a Raspberry Pi to an e-paper display and wrote a piece of code that could draw rectangles in two shades of grey, so only 48 to go! Or maybe we’ll try the Mendeley Logo next time.



For those of us who always wondered (putting my virtual hand up here) Hadoop is essentially a tool that solves big problems by breaking them down and sending smaller problems to different machines before reassembling all the answers at the end. This is done using what is basically a large cluster of computers, so Matt Thomson came up with the idea of seeing how a cluster of Raspberry Pis might work in a similar way. He only had 3 Pis available, so the cluster was very small, but still able to solve some of the problems Matt used to put it through its paces, such as – rather appropriately – calculating the value of Pi…



Ev3 was not lonely during our Hack Day, because Robert Knight also built an awesome contraption, consisting of a robotic arm that he put to work as he orchestrated the demos and presentations. “Normal People press the space key to click through a presentation, but that’s much too simple,” says Rob. “we’ve used a robot arm holding a pen connected to a Raspberry Pi listening for commands sent to it via a group chat channel using an app on my iPhone. Just need to get Siri and Google Translate in the mix for extra Rube Goldberg points.” Next time maybe!


Mendeley at ACM Recommender Systems 2013



By Mark Levy, Senior Data Scientist at Mendeley

Last week I had the pleasure of travelling to Hong Kong to give two workshop presentations at the ACM Recommender Systems conference.  The art and science of recommender systems have come some way since the first time that “users who like X also like Y” appeared on an e-commerce site on the internet, and this year’s conference attracted several hundred delegates from both industry and academia.  Despite its close association with customer satisfaction and the commercial bottom line, as a research topic Recommender Systems occupies a tiny and somewhat recherché niche within the computer science discipline of Machine Learning, which centres on the idea that if you present a computer program with enough examples of past events, it will be able to come up with a formula to make predictions about similar events in the future.  For a recommender system these events record the interaction of a user with an item, for example Alice watched Shaun of the Dead, or Kris read Thinking Fast And Slow, and the program’s predictions consist of suggested new books that Alice or Kris might like, or of other movies similar to Shaun of the Dead, and so on.  In our products these scenarios correspond to Mendeley Suggest, currently available only if you subscribe to a Pro, Plus or Max plan, and to the Related Research feature which we recently rolled out to all users in Mendeley Desktop.

One challenge for anyone trying to build a recommender system is that it’s hard to tell whether or not your predictions are going to be accurate, at least until you start making them and can see how often your users actually accept your suggestions.  As there is a huge space of possible methods to choose from – far too many to test every possibility on unsuspecting users – ideally we’d like to be able to figure how well each prediction formula (technically each mathematical model) matches reality before we get to that stage.  If and how that might be possible was a recurring theme of this year’s conference, and the subject of my first talk in Hong Kong.

Surprisingly for a field that has now seen several years of quite intense research interest and hundreds of peer-reviewed publications, most practitioners remain highly sceptical of the results reported even in their own research.  This made it particularly interesting to hear conference presentations from large tech companies such as Google, Microsoft, LinkedIn, Ebay, not to mention Chinese counterparts such as Douban, TenCent and AliBaba, which were new names to me but who also operate at colossal scale.  These organisations have both the scientific expertise to develop cutting edge methods and the opportunity to test the results on significant numbers of real users.  You might be surprised to learn quite how much sophisticated research has gone into recommending which game to play next on your XBox.

At Mendeley we use a great deal of wonderful open source software, and so we’re very happy that the work we did in the Data Science team for my other presentation at the conference also gave us a chance to give something back to the developer community in the form of mrec, a library written in the very popular Python programming library and intended to make it easier to do reproducible research on recommender systems, even if you’ll still need to test your new algorithm on real people to convince most of us that it actually works.

The Reproducibility Initiative, supported by Mendeley data, gets $1.3M to replicate key findings in cancer biology.

The Reproducibility Initiative, a project we’ve written about before, has reached a major milestone. They have been awarded $1.3M in funding from the Center for Open Science and the Laura and John Arnold Foundation to replicate 50 key findings in cancer biology. Mendeley has supported the initiative by helping to design the selection process for papers, using Mendeley readership in addition to traditional citation measures.

We try to keep ahead of the issues in research, pushing for open access and better tools for researchers, and over the past few years, from the Stapel affair in psychology to the reports from Bayer and Amgen reports of their failures to replicate most of the high-impact biomedical research they have studied in-house, reproducibility has emerged as a key issue. This comes as no surprise to us, and in fact, John Ioannidis’ paper “Why Most Published Research Findings Are False” has been one of the all-time most highly read papers on Mendeley.

Read More »

Technology and Research Mendeley Masterclass

©Tom Atkinson 2013 -
©Tom Atkinson 2013

Last month we saw another edition of the global extravaganza that is Social Media Week. This time around there were over 1000 events and 25,000 attendees in 8 cities around the globe. The theme for this year was “Open & Connected” which is pretty much a perfect fit for the Mendeley philosophy. So we thought it would be great to host an event in the London SMW Hub about how technology is changing the way we conduct and fund research, how researchers interact, discover content and share their findings, as well as how the non-academic public can get involved and make a real different through citizen science initiatives.

Our Masterclass was streamed live and proved to be one of the most popular events of the week, with hundreds of people tuning in and sending their own questions.

Mendeley Co-founder and President Jan Reichelt kicked off the series of lightning presentations by explaining how Mendeley can help researchers organise their papers, but also how it went far beyond that. “Research is an inherently social activity, and Mendeley is an environment starting with productivity going over into collaboration, and that also crucially captures the social context going on around that research.”

Rachel Greene from JoVE challenged researchers to “stop reading and start watching,” explaining how the majority of the time scientists failed to accurately replicate the findings of key studies. She believes that technologies such as the one used in their peer-reviewed Journal of Visualized Experiments are much more suited for that purpose than traditional print, and can therefore dramatically increase reproducibility and the pace of scientific discovery.

“In the past everything was recorded on paper, but current science is very digital,” says IJsbrand Jan Aalbersberg from Elsevier’s Article of the Future project, which aims to improve scientific communication in all its rich facets. “All the records are digital, all the capturing of scientific data is digital, and the communication of that information of course is also digital. However the traditional publishers have not yet adapted to that, what they usually do is flatten the multidimensional, rich research that an author has created into a two-dimensional paper of text and images.” He gave insight into some tantalizing possibilities, including the ability to run variations of some experiments – in computer science for example – within the parameters of the article itself, making it a living, evolving piece of collaborative research.

Nicolai Humphreys from The Lancet told of how the meaning of the journal’s name came from the fact that “A lancet can be an arched window to let light in and can also be a sharp surgical instrument to cut out the dross” and upon founding the journal in 1823 Thomas Wakley stated his intention that the publication should serve both those functions. Fast-forward nearly 200 years and Nicolai is part of the team that is using technology to cut out the dross and make academic publishing more dynamic and cutting edge.

Emma Cooper described the journey that took their digital amusements company Team Cooper to developing a Facebook game in conjunction with The Sainsbury Laboratory to help harness the brainpower of citizen scientists to tackle Ash Dieback disease. Quoting Dr Dan MacLean, who approached them about building the game with their data, “humans are smart and humorous, and we love games.” The key to the success of Fraxinus is the human ability to recognise patterns, and this proved really addictive with players (over 38,000 in the first month), who spend 20 minutes on average playing the game, where the average tends to be around 5-10 minutes.

That is what Robert Simpson from citizen science web portal Zooniverse calls “cognitive surplus,” which describes the vast amount of time that we collectively spend on activities such as watching TV. “The human race spends 16 years every hour playing Angry Birds every hour. There’s a lot of brainpower out there and what we try to do is take that brainpower and make it more useful to researchers.” The team at Zooniverse works with researchers to design sites that take their data and presents it into a format that will let the crowd help them to achieve their objectives. In the case of Snapshot Serengeti, for example, this meant classifying the millions of pictures taken over 2 years by camera traps in Tanzania to provide new insight into wildlife dynamics.

“These days with modern technology Citizen Science is becoming a fresh new hot subject in science,” says Margaret Gold of Citizen Cyberlab, which is leveraging the web, mobile phones and other tools and platforms to enable crowd-sourced scientific research. “We give people across the globe an interactive means to either help with the collection of data or the processing of data, pattern recognition and so forth, and all this makes a very genuine contribution towards science.”

Dr Rayna Stamboliyska, a Research Fellow and Digital Content Coordinator at the Centre for Research and Interdisciplinarity in Paris, believes that technology can be used to bring research into primary schools, and that “we can change the world many kids at a time.” In these programs, PhD fellows work with school children to develop research projects, leveraging and incorporating various technologies and social media. “This not only engages them in the STEM curricula at a young age, but it’s a really gender neutral policy, so we’re addressing the problem of having so few women in science.”

But ground-breaking research often comes across the stumbling block that is lack of funding, and this is where Liz Wald from Indiegogo believes that crowdfunding can help scientists. “it’s really about getting rid of gatekeepers, knocking down barriers and taking ideas right to the crowd,” she said as she went through a few of the projects that were crowdfunded through Indiegogo, such as Kite Patch (a patch that lets people avoid mosquito bites) and uBiome (where you sent off swabs of your bacteria to them so that they could let you know more about yourself and also help the wider project to sequence the Microbiome). The message was that people will not only fund cool and useful gadgets, but all forms of science as long as you tell a good story.

If you missed it on the day don’t worry, all the presentations are on the Mendeley YouTube Channel, so you can watch them any time and let us know what you think! There are also some cool pictures of the day available on our Flickr page, we had a great time and thanks again to all our speakers and community!




JoVE Guest Blog Post – Visualising Scientific Experiments

JoVE October blog

This is the first in a series of guest posts from JoVE, the Journal of Visualized Experiments. Each month we’ll be featuring a different peer-reviewed video article with insights from their team. Like Mendeley, JoVE is using technology to make science more open, and we were really happy to have their Director of Marketing Rachel Greene join us at Social Media Week London for the Mendeley Masterclass on How Technology is Changing Research. You can watch the video of her presentation, talking about how visualising experiments enhance reproducibility on the Mendeley YouTube channel, and as usual let us know what you think in the comments below!

MALDI-TOF MS, an Adaptable Method for Protein Characterization, Visualized in a JoVE-Chemistry Video Article

(J. Vis. Exp. (79) e50635, doi:10.3791/50635 (2013))

By Val Buntrock, Ph.D.

Journal Development Editor, JoVE

A recent video-article published in JoVE, the Journal of Visualized Experiments, by a research group at the Centre National de la Racherche Scientifique (CNRS) captures the process of analyzing intact proteins using mass spectrometry (MS). In their video article, Signor et al. demonstrate how to accurately measure large proteins using matrix assisted laser-desorption ionization time-of-flight (MALDI-TOF) MS. Often, when describing MALDI-TOF MS procedures in text, essential information is omitted, leading to poor reproducibility. Part of a new trend in publishing, this video demonstration records the subtleties and nuances of this complex technique. Employing proper technique and variable consideration, this research group identifies an intact, large protein (> 100 kDa) with high sensitivity using a small amount (0.5 pmoles) of protein sample. Using these video instructions, research groups around the globe can modify this flexible technique to identify an infinite number of large proteins.


Characterizing proteins is important for determining the current state of the protein, which has severe implications to several biological processes. The significance of proteins switching between active and inactive forms via protein kinase phosphorylation events has been recognized and applied to cellular and molecular research for several decades. Researchers have gone on to show that protein folding, as determined by phosphorylation, either exposes or protects structural motifs that serve as binding sites for effector molecules. Further, the binding between protein and effector molecule controls protein function. Therefore, the initial protein phosphorylation event regulates the activity level of the protein.

Protein activity or function plays a role in switching on or off a large number of cellular processes, such as cell communication and cell replication. As structural biologists identify a growing number of disease states related to malfunctioning protein modifications and subsequent de-regulation, understanding and identifying the differences between the two states of the protein (active or inactive) has become a priority.

Protein Characterization

A simple, fast, and common method of determining the presence or absence of phosphorylation in proteins is by determining their mass using mass spectrometry (MS). MS ionizes the molecule of interest, generating a charged species, and measures its mass-to-charge or m/z ratio.  The m/z ratio is determined by the isotopic distribution of each element present, meaning that each molecule or protein has its own unique isotopic pattern or fingerprint.

Two MS techniques are typically used to ionize heavy and labile biomolecules, such as proteins: electrospray ionization (ESI MS) and matrix assisted laser-desorption ionization time-of-flight (MALDI-TOF MS). ESI MS analysis requires dissolution of the sample in a pure solvent for direct ionization from the solution mixture. MALDI-TOF MS utilizes a co-crystallization method wherein the protein is crystallized with an ultraviolet (UV) absorbing organic species. This organic molecule is referred to as the matrix molecule or substance.

While ESI MS is more sensitive and accurate, the instrument compatible solvents or buffers typically contain significant amounts of substances, such detergents and salts, which interfere with the desired protein pattern. Additionally, ESI MS data is more difficult to interpret given that ESI MS spectra are riddled with multiple overlapping charge states from a single protein. A more gentle technique, MALDI-TOF MS produces fewer multiply-charged species, leading to a much cleaner spectra that is easier to analyze. This is especially true for larger biomolecules, such as proteins, which can fragment into numerous multiply charged species using ESI MS. For these reasons, MALDI-TOF MS is the preferred technique for protein analysis.

Optimizing MALDI-TOF MS Technique


Purity at every stage of MALDI-TOF MS analysis is crucial to obtaining the highest quality MS spectra or protein fingerprint. For this reason, Signor et al. provide detailed instructions for how to effectively 1) clean the MALDI plate that holds the matrix and protein of interest and 2) purify the matrix substances using standard recrystallization techniques. Further, they employ two different matrix systems to compare which one yields the best results for the protein of interest. A single matrix and a mixed matrix system are used, demonstrating the influence of the matrix on the resulting spectra. In their work, Signor et al. found that the mixed matrix system yielded a higher signal to noise and a greater sensitivity than the single matrix system.

Deposition Method

The deposition method is another technique variable that impacts the quality of MALDI-TOF MS results. The two most commonly employed deposition methods are the droplet and thin layer method. Using a droplet technique, a mixture of the protein analyte and matrix solution is “dropped” onto the target substrate, and the solvent is evaporated, yielding a crystalline mixture of the matrix and protein. Slightly more controlled, the thin layer technique is composed of layers of matrix sandwiching the protein analyte layer. While the droplet method suffers from poor resolution and an inability to observe larger proteins (> 100 kDa), the thin layer deposition yields a protein fingerprint for large proteins, sharper peaks, and a higher signal to noise.

As protein analysis becomes a more vital component of studying protein modifications, mastering protein characterization techniques is increasingly important. In this video-article, Signor et al. provide a detailed overview of the steps involved in utilizing MALDI-TOF MS to analyze proteins. They also provide essential considerations and modifications to guide beginners and experts alike through tailoring this powerful technique to study different target proteins. Shown in video format, the necessary level of details, such as how to properly perform multiple deposition methods, is captured and relayed to the viewer for increased transparency.





Mendeley Desktop 1.10

The next release of Mendeley Desktop is here. You can update from within the app via Help → Check for Updates or download it here.

Mendeley Desktop 1.10 features improvements in two main areas: creating citations and discovering relevant research.

Journal Abbreviations

To make it easier to generate correct citations Mendeley Desktop 1.10 will now automatically abbreviate publication titles according to the rules of the style. We’ve included a built-in set of abbreviations and rules which should cover most publications out of the box. You can however provide additional abbreviations or custom lists in cases where the built-in ones do not match the output you need.


For BibTeX users, we’re handling abbreviations by providing an option to control whether the ‘journal’ field contains the full publication title (the default) or an automatically generated abbreviation. This can be set from the BibTeX tab in the Options/Preferences dialog.


In addition to journal abbreviations, we’ve also been checking and adjusting the styles used for many Elsevier journals.

Related Research

A key goal of Mendeley is to help users discover useful content related to their research interests. On the article pages in our catalog, we display papers related to the current entry.

In this release, we’ve added a new Related Research view which provides a way to quickly find articles in our catalog that are related to the selected content in Mendeley Desktop. With Related Research you can get:

  • Instant recommendations based on specific articles – Select a single document, a group of documents or a whole folder just hit the ‘Related’ button to find research related to those papers.
  • Quick and easy import – You can add recommended papers to your library with a single click and if we have permission to do so, we’ll include the full text of the article as well.
  • Drill-down into recommendations – When you find a recommendation that interests you, select it and click ‘Related Documents’ again to view further suggestions based on that paper.

Using Related Research is simple. Just select the document(s) or folder you are interested in and click the ‘Related’ button on the toolbar:


Mendeley Desktop will then search the catalog of the world’s research that we’ve crowdsourced from users and publishers and present related papers:


The recommendations come from a mix of analysis based on the content of the selected paper and what other users with similar interests are reading. Our new recommendations service is still young and we’re working on tuning the results, so please give us feedback and let us know how we’re doing.

Other Improvements

Alongside the two main improvements this release also includes a number of important bug fixes and stability and performance improvements. See the release notes for more details.

Supported Platforms

We are planning changes to supported versions of Mac, Windows and Word in the near future. This release introduces support for the upcoming OS X 10.9 and Windows 8.1 operating systems, expected to become widely available in October. There are no changes to supported versions of Mac OS X or Word in this update but in subsequent releases we intend to retire support for OS X 10.5, Word 2008 on Mac and Word 2003 on Windows.