BEAST 2 paper is out!

21 April 2014 by Remco Bouckaert

The BEAST paper “BEAST 2: A software platform for Bayesian evolutionary analysis” by Remco Bouckaert, Joseph Heled, Denise Kuehnert, Tim Vaughan, Chieh-Hsi Wu, Dong Xie, Marc Suchard, Andrew Rambaut, and Alexei J Drummond is now available from PLOS Computational Biology 10(4), in case you want to cite BEAST 2 (which 3 people already did at the date of publication, according to Google scholar 🙂 ). PLOS Comp Bio is one journal in the series of open access PLOS publications with a knowledgeable editorial board and fast review time and publication process, so there is a lot to commend PLOS Comp Bio.

However, (disclaimer: the following is my personal opinion/experience and does not necessarily reflect any of those of my co-authors) the publication process left some things to be desired. Previously I published with the Journal of Machine Learning Research, which I would consider the ideal publisher: it is open access, has a high impact factor and there are no fees. Let’s look at some specific issues comparing PLOS and JMLR.


For JMLR, MIT provides the web server, saving few hundred dollars per year it would otherwise cost for commercial web hosting. According to an insider, editor in chief and editorial board are volunteers and webmasters is a student volunteer — note that I never noticed any chance in the lean submission and manuscript management system through the years, the system just works and does not require any development (even though it looks a bit unpolished). Like for most journals, reviewers are unpaid volunteers, so there is not costs involved there. The highest cost involved is a tax accountant, and average cost per paper is estimated at $6.50, covered by donations and MIT.

PLOS Comp Bio on the other hand charges $2300 per paper to be paid by the authors. Now, JMLR is an extreme example in that it relies on volunteers to a large extend and I appreciate not every journal could be run that way. Also, this is much lower than the more than $5000 ($8 billion revenue for 1.5 million papers) on average commercial publishers receive per paper, but still, it is a significant amount.

Paper submission

For JMLR, submitting a paper does not take a lot of effort — it is a one page form that only requires entering all authors on your paper, uploading the pdf paper and a cover letter (if you want to), and entering the title. Apparently, that is all that is required for running a high quality journal for a specialised field. Recommending reviewers is not required, since the editorial board is sufficiently knowledgeable to know the people in the field. If you need to declare conflict of interest, this can be done in the cover letter, and one does not have to go through some hoops like copying some standard phrase in a text field declaring no conflict of interest exists.

Compare that to submissions to PLOS; there are many HTML pages involved, and lots of information is required — such as conflicts of interests — before one is able to complete a submission. Just for going through the process of submitting a paper and filling in the forms one has to reserve a couple of hours.


For JMLR, a Latex style file (jmlr2e.sty) is provided that is easy to use. You prepare your file as any Latex file and have full control over how it looks like.

JMLR article format instructions are 2866 words when copying the page and counted in Word while PLOS’s is just over 7000 words, more than double that. But that is not all, there are separate guidelines for Latex document, figure and tables, and Supporting information none of which an author should be burdened with, since this is where publishers can add value, but apparently chooses not to.

Production process

JMLR author guidelines are short, to the point and contains a 8 point checklist that is helpful in submitting your paper. PLOS just insists that each author has to become an expert in publishing production processes — requiring skills such as converting between graphic file formats, embedding fonts in EPS files, scaling graphics to size, etc. None of these skills are required when using Latex with JMLR — you can just use your favourite file format and stick with that, and submit the Latex output.

I appreciate that PLOS publishes in different formats, and given the Latex source, it should be able to do so. Unfortunately, that process is broken — it does not convert characters with umlauts properly, and inserted random characters, in the case of the BEAST paper. One would think that these would be issues that should have been fixed by now. The amount of time me and my co-authors wasted on tedious typesetting issues like the above would be zero if we would have published with a system that JMRL provides.

Most authors that publish with PLOS do not use Latex, so this might be a relatively convenient process for the majority of papers submitted as Word document, but I think there is room for improvement.


I am picking out PLOS here because that is where the BEAST paper is published, but there are many other journals that come with a similarly uncomfortable production process.

Unfortunately, for most bioinformatics papers, JMLR is not an option and it would not have been appropriate for the BEAST paper. I am curious about experience with PeerJ, which appears to offer a smooth interface for submitting papers in Latex. I am not quite sure about the requirements for images though — the author guidelines (a pleasant 3000 word document) still mention the arcane EPS format.