Saturday, October 3, 2015

PeerJ vs F1000Research

A purely practical comment about point 5 in general and F1000Research price in particular. My main point is that PeerJ offers better service at lower cost (and I am not affiliated with PeerJ in any way).

Let’s take my latest paper which just got accepted in PeerJ and contrast it to how it would have worked at F1000Research

1. I submitted my draft to PeerJ PrePrints who made it available online within a day for free.  It showed up on Google Scholar about a week later.

F1000Research would take about a week and cost \$1000 as it was >2500 words.  On the other hand at this point it is typeset.

2. I solicit reviews on social media and by emailing select experts.  There is a commenting section on PeerJ PrePrints where these reviews can be added.  I got some suggestions by email but no one added comments for this particular paper.

From what I can tell the idea is much the same on F1000Research

3. I revise my manuscript and put a new version on PeerJ PrePrints with another plea for comments/reviews.  Then I submit to PeerJ.  PeerJ finds 2 reviewers for me, typesets the manuscript (after minor corrections in this case), publishes the reviews, provides a comment section for further review, and gets it indexes, for \$298 (in this case). Again, there is a comment sections where people can continue to review the manuscript and also the reviewers comments, which I choose to make public.

So, from where I stand I pay F1000Research \$1000 extra for guaranteed and immediate typesetting of a manuscript which may not get reviewed, while I pay PeerJ \$300 for guaranteed reviews of a manuscript which may not get typeset (if it is rejected). 

I couldn’t care less about the typesetting. When I deposit my preprint I consider my work published - and I can do that for free. The remaining steps are taken mainly to be able to add it on my CV under “Peer Reviewed Publication” with additional indexing as a nice bonus.  

As Gowers has shown, if you remove the typesetting this can done for \$10/paper.

This work is licensed under a Creative Commons Attribution 4.0

Sunday, September 20, 2015

Surface tension and the non-polar solvation entropy

I made a stupid sign mistake in one of my video lectures and have spent part of the weekend sorting things out in my mind.  So here is a note to self before while it is still fresh in my mind.

The non-polar free energy of solvation can be written as
$$\Delta G_{\text{np-solv}} = \gamma_{\text{np}} SASA$$
where where $SASA$ is the solvent accessible surface area.  The argument is that one of the main contributions to $\Delta G_{\text{np-solv}}$ is the energy required to make the molecular cavity in the solvent, which, for macroscopic objects, is a function of the surface tension of the liquid $\gamma$ and the surface area of the cavity.
$$\Delta G_{\text{np-solv}} \propto \gamma SASA$$
$\gamma$ is positive for water so $\Delta G_{\text{np-solv}}$ is positive in water.

So far, so good.  But this simple picture fails when considering the solvation entropy
$$ \Delta S_{\text{np-solv}} = - \left( \frac{\partial \gamma_{\text{np}}}{\partial T} \right) SASA$$
For water the bulk surface tension decreases with increasing temperature as you would expect, which suggests that $ \Delta S_{\text{np-solv}}$ is positive when in fact it is observed to be negative.

So if $ \gamma_{\text{np}}$ has anything to do with $\gamma$ this would imply that the surface tension associated with molecular-sized cavities increase with temperature.  It is not clear why that would be so and this, in part, has led Graziano to argue that, effectively, $ \gamma_{\text{np}}$ has nothing to do with $ \gamma$ but is a strictly empirical parameter.

A little more detail that ultimately doesn't shed any more light
$ \gamma_{\text{np}}$ is also positive but not equal to $\gamma$. One reason is that $\Delta G_{\text{np-solv}}$ also contains contributions from repulsion and dispersion interactions with the solute.  However, if one computes $\Delta G_{\text{np-solv}}$ from hard sphere simulations the corresponding $ \gamma_{\text{np}}$ values still does not match the $\gamma$ value for bulk water.

Tolman has argued that the surface tension depends on the curvature of the surface and suggested the following approximation
$$\gamma (R) = \gamma \left( 1 - \frac{2\delta}{R} \right)  $$
where $R$ is the cavity radius and $\delta$ is a parameter called the Tolman length.  When $R < 2\delta$
$$ \frac{\partial \gamma (R)}{\partial T} =  \left( \frac{\partial \gamma}{\partial T} \right) \left( 1 - \frac{2\delta}{R} \right)$$
will indeed be positive, but only when $\Delta G_{\text{np-solv}}$ is negative.  What is observed is positive $\Delta G_{\text{np-solv}}$ and  $\Delta S_{\text{np-solv}}$.

Ashbaugh has pointed out that a  temperature-dependent $\delta$ solves this problem but Graziano fired back that since there is no analytical form for $\delta$, $\frac{\partial \delta}{\partial T}$ is just another temperature dependent parameter, and you might as well use $\frac{\partial \gamma_{\text{np}}}{\partial T}$ as a parameter (I am paraphrasing here).

This work is licensed under a Creative Commons Attribution 4.0

Saturday, September 19, 2015

ProCS15 paper: reviews are in

2015.10.2 update: Our rebuttal can be found here.  The paper is now accepted.

The reviews of the ProCS15 paper we submitted on August 25 arrived last evening. 25 days to first decision. The verdict was "minor revisions". The editor was Freddie Salsbury, Jr (who also handled our very first PeerJ paper) and both reviewers chose sign their reviews.  Another very pleasant publishing experience with PeerJ.

Editor's comments
Both reviewers have some minor corrects to make and the second reviewer raises a point of skepticism about QM-based vs empirical estimators. A discussion addressing this would likely be of benefit to the field.

Reviewer Comments
Reviewer 1 (Xiao He)
Basic reporting
No Comments
Experimental design
No comments
Validity of the findings
No comments
Comments for the author
This manuscript is of great importance and I totally support its publication in PeerJ. The authors present an excellent and accurate chemical shift prediction program (ProCS15) based on millions of DFT calculations on simplified models. ProCS15 has extended the capability of previous ProCS program, which predicts the backbone amide proton chemical shift, to fast estimation of chemical shifts of backbone and C beta atoms in large proteins. The accuracies of chemical shifts on two proteins (namely, Ubiquitin and GB3) predicted by ProCS15 are very close to the results from fragment-based DFT calculations by Zhu et al., and Exner and co-workers. Nevertheless, the computational cost of ProCS15 is within a second. This program will be widely used in the NMR community. I only have a few minor points.

1) In the Introduction section, “RMSD observed for QM-based chemical shift predictions may, at least in part, be due to relatively small errors in the protein structures used for the predictions, and not a deficiency in the underlying method.” I agree with the first half of the statement, however, the limitation of current density functionals also contributes to the discrepancy between experiment and DFT calculations, especially for the 15N chemical shift prediction.

2) The first AF-QM/MM work is highly recommended to be cited in the paper,
He X., Wang B. and Merz K.M., Protein NMR Chemical Shift Calculations Based on the Automated Fragmentation QM/MM Approach. J. Phys. Chem. B 113, 10380 (2009)
Reviewer 2 (Dawei Li)
Basic reporting
No comments.
Experimental design
No comments
Validity of the findings
No comments
Comments for the author
This work is a direct extension of the author’s previous work on quantum based protein chemical shift calculation. The performance is comparable to other quantum based predictors but is worse than current empirical predictors. Because of this, I am still skeptical about all quantum-based predictors. Without solid cross-validation, it is very hard to argue that quantum predictors can capture subtle effect better than empirical predictors. It is true they respond more sensitively to minor structural change, but not necessary in a correct way. On the other hand, it is very useful for the whole community to have more selections that is different from previous ones. (Note that predictions from most empirical predictors are highly correlated, i.e., it won’t provide more information by switching from one to another empirical predictor.) In this context, this work should be published.

It is nice that the prediction performance can be improved a lot if applied to more realistic NMR-derived ensembles. This is expected because the experimental chemical shift of a given nucleus reflects the Boltzmann-weighted average of the 'instantaneous' chemical shifts of a large number of conformational substates that interconvert on the millisecond timescale or faster. This behavior has been discussed many times in the literature. All Ubiquitin NMR structures cited in this work are generated specifically to be a more realistic presentation of protein ensemble in solutions, except 1D3Z. 1D3Z is a traditional NMR structure model, where NMR conformer “bundle” should not be confused with a dynamic ensemble representation of the protein. In these types of NMR models, the spread of atomic positions merely provides information about the uncertainties of the atomic positions with respect to the average structure and has no direct physical meaning. The author may need to provide more comments on this in their last section titled “Comparison to experimental chemical shifts using NMR-derived ensembles”.

Saturday, September 5, 2015

Why I chose to become a subject editor for the RIO Journal

I agreed to become a subject editor on a new journal called The Research Ideas and Outcomes (RIO) Journal. When Scientific Reports asked be I declined. Here's some of the reasons why I said yes to RIO Journal, in rapidly descending order of importance.

1. The world needs a low-cost alternative to PLoS ONE*
Many people say things like "I couldn't afford to publish all my papers OA at $1350/paper" and so they publish none as OA. While PLoS ONE offers a no-questions-asked full or partial fee-waiver most people feel funny about asking for it (not me though). PeerJ and PeerJ Computer Science offer very cost effective alternatives to PLoS ONE for bio- and computer science-related papers.  For example, on average a PeerJ paper costs me about $200-300. But what about other areas? I was assured that the cost of publishing in RIO Journal would be comparable to PeerJ.  Should this prove not to be the case (the pricing is still a bit up in the air) then I'll resign as subject editor.

(*note that this implies PLoS ONE-like review criteria and use of the CC-BY license)

2. I like the idea of getting "publishing-credit" for my research proposals and other research output
Roughly speaking for every proposal I write, I write one paper less. With the current ~10% success rate I now write more proposals and, hence, fewer papers. I would like to change that because my productivity is judged in large part by my production of peer reviewed papers, and RIO Journal looks like the way to do this.

There are plenty of places where you can share your proposals (I have used figshare which even gives you a DOI) but if I can get them peer reviewed (what RIO Journal calls "validated") at RIO Journal then I can list them on my publication list and get "credit".  If RIO Journal can deliver this for $200-300 count me in.

3. All the other stuff
A. The manuscript is visible upon submission, i.e. you "automatically post your pre-print".
B. The reviews are made public and are assigned DOIs
C. Commenting is possible
D. The people behind the journal are doing this to improve science rather than making money

All these things are very nice but I am not willing to pay extra for it.

This work is licensed under a Creative Commons Attribution 4.0

Sunday, August 23, 2015

Thermodynamics for Biochemists: a YouTube textbook

As you may know I don't use textbooks in my courses anymore.  Instead I make my own video lectures and make the corresponding slides accessible.  Inspired by Engineering Mathematics: YouTube Workbook I have now organized the slides from one of my courses into a "YouTube textbook".

It's mostly in Danish but a few subsections are in English.  I am working on a fully English version where the bottleneck is re-recording the videos.  You can see the progress here.

This work is licensed under a Creative Commons Attribution 4.0

Tuesday, August 18, 2015

Writing an informed teaching statement for a university faculty position

In the US new open faculty positions are starting to be announced and ChemBark and Chemjobber are curating a list for chemistry this year.  Most of these positions will require a teaching statement. When I wrote my teaching statement back in 1996 I really didn't know what to write. You find a textbook, make some lecture notes, show up 3 hours a week and write on the blackboard, and assign some problems in the book. How do you fluff that up so it fills a page?

The main point of this blog post is that in 2015 the traditional lecture model is just one of many teaching styles you can choose and you should make an informed decision, i.e. even if you choose to lecture you now really have to argue why you choose that.  What follows is a very, very brief overview (CliffsNotes) that is mainly intended to introduce you to terms that you have not have heard of but that you really need to know in order to make an informed decision.

Alternatives to the lecture approach
The main argument against the lecture model is that students only real learn by actively doing, do the the most general umbrella term for these new approaches is active learning. One fairly popular variant of active learning is project based learning which can be combined with inquiry-based learning.  These approaches can be hard to implement for large-enrollment courses. Another, increasingly popular, variant of active learning is the flipped classroom approach. The flipped classroom is often equated with blended-learning and video lectures, but the flipped classroom approach can also be based on a textbook.

There are several variants of the flipped classroom that differ on how the "lecture" time is used.  The most basic implementation of flipped classroom is simply to use the lecture time as help sessions for homework. The flipped classroom approach can also be combined with inquiry based learning using the POGIL approach.  Perhaps the most popular variant of the flipped classroom approach is the peer instruction or "clicker" approach, which scales very nicely to very large courses.

Another interesting new idea in education (that can also be used with the standard lecture model) is specification grading which is part of a relatively new movement within higher education called competency-based learning.

Finally, here are two recent peer-reviewed studies that document improvements in learning compared to the traditional lecture approach: Active learning increases student performance in science, engineering, and mathematics and Improved Learning in a Large-Enrollment Physics Class

Some pedagogical terms and concepts
Here are some  pedagogical terms and concepts that should inform your teaching no matter which style you choose.
Cognitive load - you can learn up to 7 new things at a time
Spaced learning - it doesn’t stick until you’ve seen it 3-4 times over a period of time
Formative assessment - you learn by answering questions if you get immediate feedback
Just in case vs just in time teaching - “You’ll need to know this later” is not a good motivator

Video Lectures and Web clickers
Though not strictly required, many teachers who use the flipped classroom approach make video lectures. This can be done relatively cheaply and easily using screencasting software such as Camtasia or Screenflow together with Powerpoint. One can also make pencasts (handwritten video lectures) using, for example, iPad apps such as Explain Everything or the Livescribe pen but such pencasts often appear too slow when watched online.

Traditional clickers are increasingly being replaced by "web-clickers" such as Socrative or Poll Everywhere on smartphones and laptops

More information
Active learning: tools and tips
My flipped classroom: what I did and how I did it
Why Not Try A Scientific Approach To Science Education?
Psychological insights for improved physics teaching
Carl Wieman Science Education Initiative - Resources
Confessions of a converted lecturer (Youtube) (abbreviated version)

This work is licensed under a Creative Commons Attribution 4.0

Sunday, August 16, 2015

Finding the melting temperature by coexistence simulations

While reviewing a paper I came across the idea of finding melting temperatures by coexistence simulations.  The idea is very simple:

1. Run an $NVE$ MD simulation starting from a configuration where half the molecules are solid and the other half is liquid.
2. If the choice of $E$ is such that both phases exist after equilibrium then the average temperature will of the system will converge to the melting temperature.
3. In practice you determine $E$ by running a short $NVT$ MD simulation, where $T$ is reasonably close to the suspected melting temperature, and using the final position and velocities as initial conditions for the $NVE$ simulation.  It's probably best to use a range of temperatures.
4. If you want the melting temperature at constant $P$ run an $NPH$ MD simulation instead.

I traced the approach back as far as this paper, which also has a nice explanation of why this works:
A more direct approach (Ref) to finding the transition temperature is to avoid the nucleation problem altogether, i.e., by simulating coexisting phases and allowing the system to evolve to equilibrium. If the equilibrium system contains both solid and liquid phases, then the system will be at a melting point. This approach is suitable for both experimental and theoretical studies. Molecular dynamic (MD) techniques are particularly useful for this approach, due to the fact that total energy is conserved in conventional MD schemes. To understand how this helps the system evolve toward equilibrium, consider a system with a phase boundary. If the system as a whole is at a temperature slightly below the melting point, then some portion of the liquid phase will solidify, generating the appropriate latent heat. Because the system is closed, this heats up the system towards the melting point. Similarly, if the system is above the melting temperature, the latent heat required to melt the solid will cool the system. The pressure of the system will also tend to equilibrate; thus, the system will evolve toward an equilibrium phase. There is no difficulty in nucleating either the liquid or solid phases, as the interface assists in the nucleation for the melting or solidification process.
The paper references an earlier book chapter, which I didn't bother to get a hold of,* that might reference even earlier works.  (*yet another demonstration of why publishing original work as a book chapter is equivalent to burying it in your backyard).

This work is licensed under a Creative Commons Attribution 4.0