Refs

Author: davidlmobley

paper/basic_training.bib

@@ -878,3 +878,15 @@ @article{Kofke1993
 volume = {98},
 year = {1993}
 }
+
+@article{Shirts:2007:JPhysChemB,
+  title = {Accurate and Efficient Corrections for Missing Dispersion Interactions in Molecular Simulations.},
+  volume = {111},
+  doi = {10.1021/jp0735987},
+  number = {45},
+  journal = {J Phys Chem B},
+  author = {Shirts, Michael R and Mobley, David L and Chodera, John D and Pande, Vijay S},
+  month = nov,
+  year = {2007},
+  pages = {13052-13063},
+}

paper/basic_training.tex

@@ -338,6 +338,8 @@ \subsubsection{Books}
 \end{itemize}
 
 \subsubsection{Online resources}
+
+Several online resources have been particularly helpful to people learning this area, including:
 \begin{itemize}
 \item David Kofke's notes: \url{http://www.eng.buffalo.edu/~kofke/ce530/Lectures/lectures.html}
 \item Scott Shell's notes: \url{https://engineering.ucsb.edu/~shell/che210d/assignments.html}
@@ -373,7 +375,7 @@ \subsubsection{Key concepts}
 Because so many interactions in physical systems involve polarity, and thus significant long-range interactions that decay only slowly with distance, it is important to regard electrostatic interactions as fundamentally long-range interactions.
 Indeed, contributions to the total energy of a system from distant objects may be even more important in some cases than those from nearby objects.
 Specifically, since interactions between charges fall off as $1/r$, but the volume of space at a given separation distance increases as $r^3$, distant interactions can contribute a great deal to the energies and forces in molecular systems.
-In practice, this means that severe errors often result from neglecting electrostatic interactions beyond some cutoff distance~\cite{Leachbook, York:1993:J.Chem.Phys., Darden:1993:J.Chem.Phys., Piana:2012:PLOSONE, Sagui:1999:Annu.Rev.Biophys.Biomol.Struct.}. 
+In practice, this means that severe errors often result from neglecting electrostatic interactions beyond some cutoff distance~\cite{LeachBook, York:1993:J.Chem.Phys., Darden:1993:J.Chem.Phys., Piana:2012:PLOSONE, Sagui:1999:Annu.Rev.Biophys.Biomol.Struct.}. 
 Thus, we prefer to include \emph{all} electrostatic interactions, even out to very long range. 
 Once this is decided, it  leaves simulators with two main options, only one of which is really viable.
 First, we can simulate the actual finite (but large) system which is being studied in the lab, including its boundaries.
@@ -451,8 +453,7 @@ \subsection{Force fields}
 \label{sec:force_fields}
 
 The term ``force field'' simply refers to the included terms, particular form, and specific implementation details, including parameter values, of the 
-chosen potential energy function.
-\footnote{It is worth noting there is a occasionally a bit of ambiguity when the term ``force field'' is used.
+chosen potential energy function.\footnote{It is worth noting there is a occasionally a bit of ambiguity when the term ``force field'' is used.
 In some cases it is used to refer to a library of parameters that could be applied to assign an energy function to a specific molecular system via a parameterization process after applying some specific chemical perception like atom typing to that system~\citep{Mobley:2018:bioRxiv}.
 For example, one might speak of the AMBER ff15FB~\citep{amber15FB} protein force field, which essentially provides a recipe for assigning parameters to a protein once atom types are assigned.
 In other cases, ``force field'' is used to refer to the specifics of the potential energy function after application to a specific system --- what could also be called a ``parameterized system''. 
@@ -461,13 +462,16 @@ \subsection{Force fields}
 Most of the terms included in potential energy functions have already been detailed in Section~\ref{sec:mol_interactions}, with the most common being Coulombic, Lennard-Jones, bond, angle, and torsional (dihedral) terms. 
 Here, we very briefly describe the mathematical forms used to represent such interactions.
 
-Non-bonded interactions of the Lennard-Jones form are well-described throughout the literature (for instance see Ch. 4 of \citet{LeachBook}). 
+Non-bonded interactions of the Lennard-Jones form are well-described throughout the literature (for instance see Ch. 4 of \citet{LeachBook}); these model a short-range repulsion that scales as $1/r^{12}$ and a long-range attraction that scales as $1/r^6}$ 
 Coulombic interactions, including both short and long-range components, are described in detail elsewhere in this document.
 To represent bonded interactions, harmonic potentials are often employed. 
-The same is true for angles between three bonded atoms, but the harmonic potential  is applied with respect to the angle formed and not the distance between atoms. 
-Torsional terms are also commonly employed, usually consisting as sums of  cosines.
-More exotic potentials based on three-body intermolecular  orientations, or terms directly coupling bond lengths and bending angles are also possible. 
-Additionally, other choices of potential function are of course  acceptable, including Buckingham or Morse potentials, or the use of ``improper'' dihedral terms to enforce planarity of cyclic portions of molecules. 
+The same is true for angles between three bonded atoms, but the harmonic potential is applied with respect to the angle formed and not the distance between atoms. 
+Torsional terms are also commonly employed, usually consisting as sums of cosines, i.e. a cosine expansion.
+
+While the above are perhaps the most common potentials used, there are a variety of common variations as well.
+More exotic potentials based on three-body intermolecular orientations, or terms directly coupling bond lengths and bending angles are also possible.
+Some historic force fields also added an explicit (non-Coulombic) hydrogen bonding term, though these are less frequently used in many cases today.
+Additionally, other choices of potential function are of course acceptable, including Buckingham or Morse potentials, or the use of ``improper'' dihedral terms to enforce planarity of cyclic portions of molecules. 
 This may even include empirical corrections based on discrete binning along a particular set of degrees of freedom~\citep{mackerell2004CMAP, perez2015}, as well as applied external fields (i.e. electric fields) and force field terms describing the effect of degrees of freedom, such as solvent, that have been removed from the system via ``coarse-graining.''~\citep{sanyal2016}
 For a more in-depth discussion of common (as well as less common) force field terms, see Ch. 4 of \citet{LeachBook}, or for an in-depth review of those specific to simulating biomolecules, see \citet{Ponder2003}.
 
@@ -477,21 +481,27 @@ \subsection{Force fields}
 For this reason, one functional form may be preferred above another due to  enhanced numerical stability or simplicity of implementation, even though it is not as faithful to the underlying physics.
 In this regard, force field selection is a form of selecting a model -- one should carefully weigh the virtues of accuracy and convenience or speed, and be ever-conscious of the limitations introduced by this decision (for instance, see \citet{Becker2013}).
 It is also important to know that most MD simulation engines only support a subset of functional forms.
-For those forms that are supported, the user manuals of these software packages are often excellent resources for learning more about the rationale and limitations of different potential energy functions and terms (see Part II of Amber reference manuals\citep{AmberManual} and Ch. 4 of the reference manual for GROMACS \citep{GROMACSManual}).
-\todo[inline, color={green!20}]{JIM: I just referenced the manuals that I'm familiar with... additional citations would be appreciated and seem appropriate. Right?}
+For those forms that are supported, the user manuals of these software packages are often excellent resources for learning more about the rationale and limitations of different potential energy functions and terms (e.g. see Part II of Amber reference manuals\citep{AmberManual} and Ch. 4 of the reference manual for GROMACS \citep{GROMACSManual}).
+
+For practical purposes, most beginning users will not be fitting a force field or choosing a functional form, but will instead be using an existing force field that already relies on a particular functional form and is available in their simulation package of choice, so for such users it is more important to know how the functional form represents the different interactions involved than to necessarily be able to justify why that particular functional form was chosen.
 
-Many examples of force fields abound in the literature, in fact too many to provide even a representative sample or list of citations, as most force fields are specifically developed for particular systems under study.
+Many examples of force fields abound in the literature --- in fact, too many to provide even a representative sample or list of citations, as most force fields are specifically developed for particular systems or categories of systems under study.
 However, reviews are available describing and comparing force fields for biomolecular simulations~\citep{Ponder2003, Riniker2018}, solid, covalently-bonded materials~\citep{Mishra2017}, polarizable potentials~\citep{Lopes2009}, and models of water~\citep{Onufriev2018, Vega2011}, to name just a few. 
-Many force fields are open-source and parameter file libraries may be found through the citations in the resources above or are often provided within a distributed molecular dynamics package. 
+Many force fields are open-source and parameter file libraries may be found through the citations in the resources above or are often distributed with molecular simulation packages. 
 Limited databases of force fields also exist, most notably for simulations of solid materials where interatomic potentials display a much wider array of mathematical forms~\citep{openKIM, IPRnist}.
 
-For almost all force fields, many versions, variants, and modifications exist.
-Each force field version not only includes specific functional forms to describe intermolecular potentials, but also specifies the parameters governing such interactions.
+Specification of a force field involves not just a choice of functional form, but the details of the specific parameters for all of the interacting particles which will be considered --- that is, the specific parameters governing the interactions as specified by the functional form.
 Parameters are usually specific to certain types of atoms, bonds, molecules, etc., and include point charges on atoms if electrostatic terms are in use. 
-A subtle and often overlooked point involving force field implementation concerns the consistent treatment of constraints, cut-offs or other simulation settings that impact the potential energy function.
-To exactly replicate a described parametrization of a force field, constraints should be specified consistently in addition to usage of the same intermolecular cut-offs as used in the original paper.
-The choice of how to apply a cutoff, such as through direct truncation, shifting of the potential energy function, or through the use of switching functions, should be maintained if identical matches to properties of interest are desired.
+
+Some choices which are often considered auxiliary actually comprise part of the choice of the force field or interaction model.
+Specifically, settings such as the use of constraints, the treatment of cut-offs and other simulation settings affect the final energies and forces which are applied to the system.
+Thus, to replicate a particular force field as described previously, such settings should be matched to prior work such as the work which parameterized the force field.
+The choice of how to apply a cutoff, such as through direct truncation, shifting of the potential energy function, or through the use of switching functions, should be maintained if identical matches to prior work computing the properties of interest are desired.
 This is especially important for the purposes of free energy calculations, where the potential energy itself is recorded.
+However, force fields are in some cases slow to adapt to changes in protocol, so current best practices seem to suggest that lattice-sum electrostatics should be used for Coulomb electrostatics in condensed phase systems, even if the chosen force field was fitted with cutoff electrostatics, and in many cases long-range dispersion corrections should be applied to the energy and pressure to account for truncated Lennard-Jones interactions~\cite{Shirts:2007:JPhysChemB}.
+\todo[inline, color={yellow!20}]{DLM: Probably need more cites here.}
+
+For almost all force fields, many versions, variants, and modifications exist, so if you are using a literature force field or one distributed with your simulation package of choice, it is important to pay particular attention (and make note of) exactly what version you are using and how you obtained it so you will be able to accurately detail this in any subsequent publications.
 
 As clearly described in \citet{Becker2013}, it is of paramount importance to understand the capabilities and limitations of various force field models that may seem appropriate for one's work. 
 Depending on the physics being simulated and the computational resources at hand, no force field in the literature may provide results that accurately reproduce experiment.