Re: [AMBER] TIP4P-Ew vs TIP3 vs TIP5

From: Saeed Izadi <>
Date: Wed, 18 Jan 2017 13:17:45 -0800

 I don’t have a negative atittude against 4-site models, in fact I am a fan
of them (especially the more recent ones). 4-site models are designed to be
more accurate than 3-site models in bulk properties and electrostatics,
however since protein force field parameters are historically optimized
against TIP3P, in many cases the use of TIP3P in biomolecular simulations
still leads to more realistic results, even though TIP3P is really bad in
bulk properties. That said, there are still many cases where the benefit of
a more accurate water model outweighs the benefit of using a water model
which is more compatible with the current force fields: for example, when
the extended structure or partially unfolded parts in structures are of
interest, the energy costs due to solvent-solvent interactions become
significantly more important than the case where a compact structure is
being studied. Therefore, a water model which may be more accurate in
representing bulk properties should be more suitable for simulating
extended structures such as intrinsically disordered proteins or RNAs which
are floppy and extended.. so at the moment, using the current force fields,
the choice of water model really depends on the structure you are
interested to simulate (maybe TIP3P for compact states and recent 4-site
models for extended states)..this might change if the future force fields
be parametrized using more accurate water models.. A safer bet at the
moment might be to run one simulation using TIP3P and one using one of the
recent 4-site models to see if you get any difference..


On Wed, Jan 18, 2017 at 3:13 AM, DmitrySuplatov <> wrote:

> Dear Saeed,
> thank you for the detailed comments.
> My specific situation is that I want to simulate conformational
> rearrangements in protein structures (e.g. movement of loops) in water at
> 300K on a microsecond scale (500 - 1000 ns) using classical MD. I expect
> that some conformational rearrangements should happen but do not want to
> accelerate them by increasing the temperature, so instead I calculate
> longer trajectories.
> From your reply I conclude that TIP3P "... may slightly speedup the
> conformational sampling ..." - so this is exactly what I need, right?
> However in the previous sentence you say that "...TIP3P seems to over
> stabilize...the native state" which seems to contradict with the first
> statement...
> I am also confused about your overall negative attitude towards the 4-site
> waters. My naive expectation was that 4-site models should generally be
> always better that 3-site models and thus worth the processing overhead.
> Otherwise, why would anybody use them, right?
> If TIP4P-Ew model is so bad than why was it used in [Shaw D. E. Science
> 2010 , 330, 341] and also in this Amber tutorial -
> ???
> Finally, is there some general guideline on how-to choose a water model
> for a particular task.
> Thank you for your time.
> Dmitry
> On 01/18/2017 08:13 AM, Saeed Izadi wrote:
> My hunch is that for many simulations at 300K TIP3P with current
> force-fields is better than TIP4P-Ew. Small molecule solvation is a
> documented example. On the other hand, TIP3P does not get the structure of
> water right, TIP4P-Ew does a better job.
> For #1, I suspect that TIP3P is still better if the goal is to simulate
> the native state, as TIP3P seems to over stabilize it. In addition,TIP3P's
> self-diffusion is too times higher than that of TIP4P-Ew (and experimental
> value), which may slightly speedup the conformational sampling..
> #2: 373K is dicey, as TIP3P temperature dependence is bad. On the other
> hand, others have used TIP3P at high temp. as well. I am not sure if
> TIP4P-Ew is better.
> #3. Newer 4-site waters may be worth the extra expense, but probably not
> TIP4P-Ew.
> #4. At the moment "No.". Not for biological simulations.
> Also, in cases where electrostatic interactions are important (e.g.
> RNA/DNA simulations), models with more accurate electrostatics perform
> better. Note that the extra point charges do not always guarantee better
> electrostatics for the model: TIP4PEw represents the electrostatics of
> water better than TIP3P, but TIP5P is less accurate in that respect. More
> recent 3-site models can be more accurate than older 4-site models (TIP4P);
> newer 4-site models are probably most accurate in terms of electrostatics.
> Hope that helps..
> best,
> Saeed
> On Tue, Jan 17, 2017 at 7:21 AM, DmitryASuplatov <>
> wrote:
>> Dear Amber users,
>> I am using TIP4P-Ew water model for a "general-purpose" protein-in-water
>> simulations at 300K with PME electrostatics and FF14SB field.
>> Please provide your best guess to the following questions:
>> 1/ Is this (TIP4P-Ew) a good choice of a water model?
>> 2/ Would it still be a good choice of a water model if I want to
>> simulate at 373K?
>> 3/ Would you say that 4-site water is better than 3-site water (e.g.,
>> TIP3P) and worth the extra computing power?
>> 4/ Would you say that upgrading to the 5-site water would do a better
>> job (~ provide more meaningful biological predictions of a protein
>> behaviour in a water solution)?
>> Thank you.
>> Dmitry
>> _______________________________________________
>> AMBER mailing list
AMBER mailing list
Received on Wed Jan 18 2017 - 13:30:02 PST
Custom Search