Friday, September 9, 2022

Current status of the project

18.5.2022 NMRlipids databank: Quality evaluation post was published.
20.4.2022 New yaml format of mapping files post was published.
8.4.2022 The NMRlipids summer school 2022 (postponed NMRlipids winter school 2021) will be organized on 1.-3.6.2022 in Helsinki. More details and registration available from the webpage
4.3.2022 We will have a Zoom meeting about the NMRlipids VI project on 10th of March at 15:00 CET. For the Zoom invitation, send email to Batuhan Kav. See the recent post for more details.
22.12.2021 The slides about the current status of the NMRlipids databank presented in the online meeting on 17.12.2021 are available from here.
9.12.2021 Online meeting on the development of NMRlipids databank and project will take place on 17.12.2021. Program and more details available from here. For Zoom invitation, send email to samuli.ollila (@)
8.12.2021 Due to the current COVID-19 regulations of the venue, we had to postpone the NMRlipids winterschool 2021 to the first half of the next year. Nevertheless, there will be a developers meeting online in December. More information will follow when we have the new dates available.
22.10.2021 NMRlipids winterschool 2021 will be organized on 15.-17.12.2021 in Helsinki with the option for online participation. Aim of the school is to give introduction for the automatic and flexible analyses over hundreds of lipid bilayer simulations using the NMRlipids databank. Registration and further information from this link.
24.9.2021 NMRlipids20 workshop outcomes post was published

18.8.2021 NMRlipids IVb manuscript with the updated title, Inverse Conformational Selection in Lipid–Protein Binding, is now published in the Journal of American Chemical Society.

19.2.2021 Online meeting on the NMRlipids datanbank and the NMRlipids VI project on polarizable force fields will take place on 26.2.2021 in Zoom. If you are interested to join, but have not received email invitation, please send me email. The schedule of the meeting is available in here.
23.12.2020 NMRlipids VI: First results from polarizable Charmm-Drude force field post written by Batuhan Kav was published.
3.7.2020 Online meeting about the NMRlipids databank post summarizing the outcomes of the meeting was published.

22.6.2020 We are having a NMRlipids online meeting about the development of the NMRlipids databank at 16.00 CET on Monday 29th of June. After a short presentation by Samuli Ollila on the current status, we will discuss on urgent topics related to the databank development. This discussion was originally planned to take place in the postponed NMRlipids20 meeting. The meeting is open for everyone. I will share a link by email to NMRlipids contributors before the meeting. If you are not a contributor yet, but want to join to the meeting, please send me email.

7.4.2020 NMRlipids VI: Polarizable force fields post and beta version of the new data contribution script were published. 

1.4.2020 Due to the COVID-19 pandemic situation, the NMRlipids20 workshop is postponed. The workshop will be organized as soon as possible after the pandemic is over.

5.2.2020 The NMRlipids20 workshop takes place on 13-15th of May in Prague, Czech Republic. Drop us an email if you are interested to join.

2.10.2019 the NMRlipids IV manuscript about PS lipids is now accepted to be published in the Journal of Physical Chemistry B

12.9.2019 Revised version of the NMRlipids IV manuscript about PS lipids is now submitted to the Journal of Physical Chemistry B

26.6.2019 The NMRlipids IV manuscript about PS lipids is now submitted to the Journal of Physical Chemistry B

31.5.2019 The first annual NMRlipids workshop post was published

23.4.2019 NMRlipids IVb: Assembling the PE & PG results post was published

22.11.2018 The first annual NMRlipids workshop is coming!

NMRlipids workshop 2019 May 15th to 17th

20.9.2018 Homepages of my group are now published. Check and share also the announcement of open Ph.D. student position.

20.9.2018 NMRlipids III: Quantitative measure for the force field quality needed post was published

18.9.2018 NMRlipids IV: Challenges in evaluating counterion binding affinity to PS bilayers post was published

15.9.2018 Google has launched a new Dataset search engine. It seems to find the data from NMRlipids project very well.

13.9.2018 Poster presented about NMRlipids IV project in Tiny Lip­ids With Grand Func­tions workshop in Helsinki, Fin­land, 19 - 22 Au­gust 2018:

4.7.2018 A lot of data has been contributed to the NMRlipids III and IV projects. Especially the NMRlipids III project is delayed because the main focus has recently been in the ion-membrane interactions. Currently the first priority is to finish the manuscript about PS lipids from NMRlipids IV, the second to finish the manuscript about lipid-cholesterol interaction from NMRlipids III, and the third to progress the manuscript about PE and PG lipids.

4.7.2018 NMRlipids IV: First draft of the manuscript about PS lipids post was published.

3.5.2018 Samuli Ollila received a academy research fellow position from the academy of Finland for five years. The research plan includes the development of the NMRlipids project.

13.4.2018 PS-headgroup order parameter comparison now also shows results for Amber Lipid 17:

12.4.2018 New NMRlipids-related publication: Accurate Binding of Sodium and Calcium to a POPC Bilayer by Effective Inclusion of Electronic Polarization post was published

30.1.2018 Database of the NMRlipids simulations and experiments post was published.

22.12.2017 NMRlipids IV: Current status and reorganization of the manuscript post was published

8.12.2017 Results from CHARMM36 simulation with cationic surfactants was added to Quantifying the effect of bound charge on headgroup order parameters post.

27.7.2017 Quantifying the effect of bound charge on headgroup order parameters post was published.

31.3.2017 NMRlipids III: Preliminary version of the manuscript post is published.

9.3.2017 NMRlipids IV: Headgroup & glycerol backbone structures, and cation binding in bilayers with PE, PG and PS lipids post is published. Almost any kind of simulations of these lipids in bilayers would be useful at this stage.

15.2.2017 My activity in NMRlipids project has been low during the last months due to other commitments. However, I have now again possibility to advance NMRlipids III and IV projects (updates will follow soon). We have also published a blog post about the future of NMRlipids project.

29.11.2016 NMRlipids project will be presented in PHOS16 Conference (Philosophy and History of Open Science) held in Helsinki on 31.11.-1.12.2016. There should be also live stream available.

12.11.2016 NMRlipids II manuscript Molecular electrometer and binding of cations to phospholipid bilayers accepted for publication in Physical Chemistry Chemical Physics, and the preprint is available on the journal web page.

16.10.2016 Zenodo has been updated as described in their news page. There are a lot of improvements but this one is probably the most important for us: "The current 2GB per file limit is removed, in favour of a 50GB per dataset limit". This means that we do not have to split the trajectories in 2GB pieces anymore.

7.10.2016 The final version of NMRlipids II manuscript (lipid-ion interactions) submitted to Physical Chemistry Chemical Physics.

9.9.2016 NMRlipids II manuscript (lipid-ion interactions) "accepted for publication after revisions" to Physical Chemistry Chemical Physics.

13.7.2016 NMRlipids II manuscript (lipid-ion interactions) has been now submitted to Physical Chemistry Chemical Physics.

1.7.2016 NMRlipids III: Preliminary observations post was published.

30.5.2016 Toward submission of NMRlipids II publication (lipid-ion interactions) (2) post was published.

20.5.2016 The new data delivered for NMRlipids II project raised a question about the order parameter responses on bound charges in CHARMM36 model. If you have CHARMM36 simulation data of PC bilayer with known amount of charged amphiphiles and you are willing to share it for the project, please let us know.

24.2.2016  Our goal from the beginning has been to immediately publish all the scientific content related to the project. One relevant part of the content are discussions between reviewers and authors during the peer review process. We have now published two peer reviewed articles: NMRlipids I and NMRlipids V. In both cases we have asked from the editor if we can publish also the reviewers' comments since everything else is public. As expected, in the NMRlipids I case the Journal of Physical Chemistry staff replied that this is not possible. However, editorial and publishing teams of BBA Membranes'  were positive about publishing the referees' comments in the case of NMRlipids V publication. Both referees were also sympathetic to the idea. However, one of them declined stating permission to make comments available should be asked a priori, at the same time referees are invited to review a paper. This is an important learning point from this experience.

25.1.2016 The review written in the NMRlipids V project has been now accepted to be published in BBA - Biomembranes and is available also from their webpage.

19.1.2016  Does the glycerol backbone structure depend on initial structure? post was published.

21.12.2015 Towards submission of NMRlipids II publication (lipid-ion interactions) post was published.
24.11.2015 The review written in the NMRlipids V project has been now submitted.

29.10.2015 The NMRlipids I publication is already available also through the journal website.

28.10.2015 The first manuscript (NMRlipids I) based on the data and discussions presented through this blog is now accepted to be published in the Journal of Physical Chemistry B. We thank all the contributors and followers for courage to participate this project.

13.10.2015 We have received a new revision request for the first manuscript (NMRLipids I project). The first version of the reply is already in GitHub. There were essentially no new comments compared to the first revision round so I will not make a new post for this. If you have comments, you can comment the Revision requested for the first manuscript post or GitHub. If there will not be objections I will submit the revision on Friday this week (16.10.2015).

28.9.2015 NMRLipids V project: Review about validations of membrane MD simulations was published. This is a project to write an invited review on a topic strongly related to the blog content.

28.9.2015 The title of the blog has been changed to "The NMRlipids project: Open Collaboration to understand lipid systems in atomistic resolution".

24.9.2015 The NMRLipids project will be discussed in Mindtreck 2015 conference in Tampere. At least one of the sessions may be live streamed, see the facebook event.

22.8.2015 The revised version of the first manuscript is now submitted.

20.7.2015 Revision requested for the first manuscript post was published.

6.7.2015 About page describing the different subprojects and Workflow page suggesting new workflow for these projects are now published.

26.5.2015 The first manuscript produced in this blog was considered to be
"primarily directed toward an audience of specialists doing closely related work and that lack a clear description of impact on the broader field of chemistry" by the editor of the Journal of American Chemical Society and it was rejected without peer review process. Thus, the manuscript has been now submitted to the Journal of Physical Chemistry (another journal ran by american chemical society).

15.5.2015 The first manuscript produced in this blog is now submitted to the Journal of American Chemical Society.

12.5.2015 The first manuscript produced in this blog will be submitted to the Journal of American Chemical Society by the end of this weeḱ.

25.3.2015 Mapping scheme for lipid atom names for universal analysis scripts post was published.

17.3.2015  Towards first submission to journal (2) post was published.

9.3.2015 Current and future activity post was published.

6.3.2015 Samuli will talk about this project in the event organized by the Open Knowledge Finland (OKFFI) on 10.3.2015 in University of Helsinki. There will also live stream from the event through this link (user: video, pw: video)

6.2.2015  The first draft of the ion-lipid interaction manuscript was published.

16.1.2015 Towards first submission to journal post was published.

16.1.2015 The current version of the new manuscript is now updated to arXiv There will be soon a new post about the further proceeding.

23.12.2014  New version of the manuscript (2)  post was published.

21.11.2014 New manuscript written on the results reported in this blog is available for commenting: New version of the manuscript. The manuscript covers only the results for fully hydrated bilayers, effect of dehydration and effect of cholesterol. A separate manuscript will be written about ion-lipid interactions.

18.11.2014 New manuscript written about the results reported in this blog will be made available for commenting on Friday 21th of November.

12.11.2014 The post About glycerol conformations is now updated. The incorrect stereospecifity in GAFFlipid for g\(_1\) segment was due to the intial structure downloaded from lipidbook, not due to the GAFFlipid force field. The updated figure with the results:

10.10.2014 Together with Hubert Santuz we have started a GitHub organization It contains a repository: The idea is to collect all the relevant files related to the project there. There are already some files and there will be more. If you are familiar with git you can add your files by making a pull request. If you are not familiar you can also make as previously (add a link to a comment) and ask us to add it into the GitHub. Downloading the data should be straightforward without any understanding about the git system. It took a couple of hours for me to get familiar with the git system. The time was well spent and I recommend it to everyone.

7.10.2014 We have added a new page called Data Contributions as an attempt to arrange the discussion. The idea is that all the new data would be sent by commenting the Data Contibutions page. Yet, let us keep the other comments under each separate post.

1.9.2014  The post About glycerol conformations was published.

20.8.2014 Presentations describing the nmrlipids project in the International Workshop on Biomembranes - From Fundamentals to Applications were posted.

19.5.2014 The post Towards a new version of the manuscript was published.

13.5.2014 To Do List has been added as a page in the top panel.

2.5.2014  The post Response of headgroup and glycerol order parameters to changing conditions: Results, reviewing the current results for the responses of the headgroup and glycerol order parameters to the changing conditions, was published.

29.4.2014 The R/S hydrogen labeling was wrong for MacRog in the previous plot. The correct one was reported by Matti Javanainen. Here is the new plot:
Now also the MacRog is in good agreement with experiments, in addition to CHARMM.

24.4.2014 Based on discussions with Antti Lamberg and Patrick Fuchs we have now plotted the results with the sign, and the R/S hydrogen labeling

It seems that the CHARMM36 results are in the best agreement with experiments. (However, the R/S hydrogen labeling in MacRog has to be still confirmed).

16.4.2014 Patrick wrote a comment on how to tell R and S and hydrogens apart.

11.4.2014 The lipid forcefield comparison at full hydration updated—now contains results for 12 force fields.

10.4.2014 The post On the signs of the order parameters was published.

10.4.2014 We have added a page containing information about the authors of the project (see the top panel).

31.3.2014 The new version of order parameter calculation script is now available at
It will now calculate also the sign. Also the *hdb file to protonate the Berger lipids with Gromacs g_protonate tool is now available. Note that there was a bug in the script shared in the original figshare package: It takes only the first 75 lipids in to account. Thus, if you have used it for the larger systems you have not taken all the available statistics into account. For my own Berger results, this makes a very small difference though. It would be very useful if someone would make a tool which would directly calculate the order parameters from the Gromacs *trr file.

14.3.2014 The lipid force field comparison at full hydration was published.

9.3.2014 Antti demonstrated that it is possible to get a very good agreement with the experimentally measured order parameters by simply sampling a large set of randomly modified dihedral potentials, choosing the most promising ones, and repeating this randomised refinement a few times.

25.2.2014 This is our new front page: A simple list the most relevant events, ordered by date. Its purpose is to help you keep up with what is happening on the blog—in posts as well as in comments.

25.2.2014 Blog post discussing the accuracy of order parameter measurements was published.

16.2.2014  Samuli gave a presentation related to the nmrlipids-project at the Biophysical Society meeting.

13.2.2014 The first attempt to modify the Berger dihedral parameters was reported with a preliminary conclusion that removing all dihedral potentials improved the choline- but impaired the g1 order parameters.

12.2.2014  Our current knowledge of the behaviour as a function of dehydration gathered into a single plot.

23.1.2014  Our current knowledge of the behaviour as a function of ion concentration gathered into a single plot.

23.1.2014  Our current knowledge of the behaviour as a function of cholesterol content gathered into a single plot.

21.1.2014 Our current knowledge of the full hydration behaviour gathered into a single plot.

10.12.2013 Patrick filed a Redmine Bug about reaction field simulations with Gromacs 4.0.7 not being reproducible with 4.5.3., which he commented first here on Oct 25th.

29.10.2013 Samuli wrote a guest post to the MARTINI group blog: PN vector orientation not a good measure for evaluating phospholipid force field performance, use head group order parameters instead.

2.10.2013 The first results were shortly reviewed and some short term goals were set in a new blog post.

13.9.2013 The first comment and the first contribution.

10.9.2013 A post discussing the motivation for the project:
and the first three scientific posts were published:

9.9.2013 The first version of the manuscript was published.

11.7.2013 The policy for publication credits was published.

3.7.2013 The was opened with a post that stated our aim.

28.6.2013 The project was first time publicly discussed in a presentation at the Biological membranes: challenges in simulations and experiments -meeting in Paris.

NMRlipids III: Including lipid lateral diffusion

The NMRlipids III project was originally started to understand the discrepancies in form factors between simulations and experiments for cholesterol-containing membranes reported by Peter Heftberger et al. After experimental form factors for a series of POPC/cholesterol mixtures were reported by the same authors, the aim was to evaluate different simulations against experimental scattering and NMR data in order to find the parameters that would best capture lipid–cholesterol interactions.

Thanks to many contributors, a set of force fields is compared against experimental data in the current manuscript. However, the conclusions from the data are complicated because quality ranking of these simulations is not straightforward. Therefore the NMRlipids III project has been on hiatus for some time.

In the meanwhile, the Slipids model has been updated, Lipid17 and Slipids support has been added to CHARMM-GUI, and quality evaluation protocol has been quantified in the NMRlipids Databank. Because the popularity of Lipid17 and Slipids in cholesterol-containing membrane simulations is expected to increase due to the CHARMM-GUI support, we believe that the NMRlipids III project is still highly relevant and should be completed.

Moreover, in another project, Matti Javanainen analyzed lipid lateral diffusion coefficients in POPC/cholesterol mixtures simulated with CHARMM36. After applying the periodic-boundary-condition (PBC) correction by Vögele et al., the lateral diffusion coefficients agreed neither quantitatively nor qualitatively with those extracted from NMR experiments. Therefore, Matti extended the comparison to other force fields, namely Lipid17, Slipids-2020, and MacRog, and observed clear differences between their behavior. We believe that these data would nicely complement the structural evaluation against form factors and order parameters in the NMRlipids III project.

Current plan is that the NMRlipids III publication would contain the structural evaluation of POPC/cholesterol mixtures against scattering form factors and order parameters for the systems that are available from the NMRlipids Databank, and comparison of selected force fields against experimental lateral diffusion coefficients of lipids. New simulations with multiple cholesterol concentrations and box sizes, required for the PBC-correction, are already available in Zenodo for CHARMM36, Slipids-2020, Lipid17, and MacRog. These data will be soon added into the NMRlipids Databank. Also the diffusion analyses have already been performed.

We aim to draft the new version of the manuscript during autumn, after which we are looking for feedback from all contributors. Especially ideas on analyses that could explain the differences in diffusion/viscosity as a function of cholesterol between the different models would be helpful.

Because the analysis of lipid lateral diffusion coefficients will be a crucial part of the new manuscript, both Matti and Samuli will be corresponding authors, of the NMRlipids III publication. Otherwise the authorship policy remains as previously.

NMRlipids Databank: Form factor quality evaluation update

Followed by the discussion in the latest post about quality evaluation, I plotted the form factors and order parameters as a function of the simulation box size from a dataset where box size was systematically varied with different amounts of cholesterol. The results shown in Fig. 1 reveal that the order parameters and the form factor zero-points do not depend on the simulation box size; however, the lobe heights in form factors decrease with increasing simulation box size.

Fig 1: Form factors and order parameters from simulations with different number of lipids and cholesterol concentrations. Simulation data from Figure also available in here.

This means that if the quality of form factors is defined as in the latest post, i.e. using the equation from the SIMtoEXP program, it will depend on the simulation box size. Therefore, I have updated the form factor quality measure in the NMRlipids Databank to be 'the distance between the first minimum in the simulated and experimental form factors'. The 'first minimum' refers here to the first minimum found in the experimental data, which has always a q-value above 0.1Å-1.

The first minimum is well correlated with the membrane thickness, as shown in the Fig. 2C (this figure is also updated in the manuscript about the NMRlipids Databank). The second minimum of the form factor would correlate even better with both the area per lipid and the thickness (see current Fig. S1 of the manuscript), but due to noise it is hard to detect automatically from some experimental data sets. Therefore, I have decided to use only the first minimum.


Fig 2: A) Best simulations ranked based on overall order parameter quality. B) Best simulations ranked based on the overall order parameter quality of POPE lipid. C)-E) Evaluation against experimental data exemplified for a simulation with the best overall order parameter quality (C), the best quality for POPE lipid (D), and the headgroup quality for POPE (E). F) Scatter plots and Pearson correlation coefficients for the area per lipid, thickness, and experimental observables. All correlation coefficients have p-value below 0.001. More correlations shown in here.

Because order parameters did not depend on simulation box size in Fig. 1,
their quality evaluation is not changed.

Wednesday, May 18, 2022

NMRlipids databank: Quality evaluation

Defining a quantitative quality measure for lipid bilayer simulations has been one of the goals of the NMRlipids project since the beginning. Such measure is highly useful when selecting the best force field for a specific application, and for improving force field parameters, particularly with automated procedures. Based on literature review and results of the NMRlipids Project, summarized in the NMRlipids V publication, we have concluded that the C-H bond order parameters from NMR can be used to evaluate the conformational ensembles of individual lipids, and the x-ray scattering form factors can be used to evaluate the lipid bilayer dimensions. Based on the work in the NMRlipids workshops in Berlin (2019) and Prague (2021), we have now written a code that evaluates the quality of simulations in the NMRlipids Databank. The key ideas and results of the quality evaluation are described in this post. More details and results can be found from the NMRlipids Databank manuscript and from GitHub.

Results. The order parameter quality of 58 simulations and the form factor quality of 99 simulations have so far been evaluated in the NMRlipids Databank. Figure 1a shows the results for the 13 best simulations according to the overall order parameter quality; and Figure 1b shows comparison between simulations and experiments for the best simulations concerning the overall order parameters (left column), the headgroup order parameters (middle column), and the form factor (right column). Results for all ranked simulations ordered in various ways are available on GitHub.

Figure 1: a) The best 13 simulations currently in the NMRlipids Databank according to the overall order parameter quality. b) Comparison of x-ray scattering form factors and C-H bond order parameters between simulations and experiments demonstrated for the simulations giving the best qualities in the overall order parameters (left column), for the headgroup order parameters (middle), for the and x-ray scattering form factor (right).

Conformational ensembles evaluated against C-H bond order parameters from NMR. After the workshop in Prague, our idea was to define the poorness Š of each order parameter as Š=-log(P); here P is the probability mass within the experimental error for a normal distribution, whose mean is the order parameter from the simulation, and whose standard deviation is the standard error of the mean from the simulation. However, when testing this definition of Š on the simulations in the NMRlipids Databank, it turned out that the probability of the simulated order parameters to locate within experimental errors was often below the numerical accuracy of computers. To avoid such numerical instability, we decided to use the first order Student’s t-distribution instead, and calculate the probability from the equation\begin{equation}  P = f \left( \frac{S_{\rm CH} - (S_{\rm exp}+\Delta S_{\rm exp})}{s/\sqrt{n}} \right) - f \left( \frac{S_{\rm CH} - (S_{\rm exp}-\Delta S_{\rm exp})}{s/\sqrt{n}} \right),
\end{equation}where f(t) is the first order Student's t-distribution, s is the variance of the order parameter SCH calculated over individual lipids and n is the number of lipids in the simulation. Because Student's t-distribution has heavier tails than normal distribution, even order parameters far from experiments have distinguishable non-zero probabilities. Therefore, the logarithm used to define the poorness Š is not needed, and we report the qualities directly as probabilities. However, it should be noted that using the first order Student's t-distribution instead of the normal distribution slightly underestimates the statistical accuracy of order parameters calculated from simulations.

In order to rank simulations based on headgroup, acyl chain, or individual lipid qualities, the average probabilities can be calculated over lipid fragments and types. For more details see the NMRlipids Databank manuscript.

Lipid bilayer dimensions evaluated against x-ray scattering form factor. The qualities of form factors in simulations are evaluated as in the SIMtoEXP program \begin{equation}
    \chi^2 = \frac{\sqrt{\sum_{i=1}^{N_q}(|F_s(q_i)|-k_e|F_e(q_i)|)^2/(\Delta F_e(q_i))^2}}{\sqrt{N_q-1}},
where Fs is the form factor from simulation and Fe from experiment, the summation goes over the experimentally available Nq points, and \begin{equation}
    k_e = \frac{\sum_{i=1}^{N_q} \frac{|F_s(q_i)||F_e(q_i)|}{(\Delta F_e(q_i))^2}}{\sum_{i=1}^{N_q} \frac{|F_e(q_i)|^2}{(\Delta F_e(q_i))^2}}.
\end{equation}It should be noted that in this evaluation the simulation uncertainty is not accounted for in any way.

Monday, May 9, 2022

NMRlipids databank: Current status and structure

After the latest NMRlipids publication, the main focus has been in the development of the NMRlipids databank that would enable automatic analyses over all the data contributed to the project. After the original idea and preliminary version, the development has been facilitated by several meetings (Berlin, online I, online II, Prague, online III). The upcoming meeting in Helsinki on 1.-3.6.2022 will contain also educative parts for using the NMRlipids databank.

The databank is now essentially functional and preparation of the first manuscript has been started. The overlay structure and current content of the databank are illustrated in figure 1. The structure of the databank is discussed more detailed below. Quality evaluation and preliminary results will be discussed in the upcoming posts.

Figure 1: a) Structure of an overlay databank. More detailed structure of the layer 2 in the NMRlipids databank is illustrated in figure 2. b) Distribution of the lengths of the trajectories, total number of trajectories and total lenght of the simulations in the NMRlipids databank. c) Distribution of lipids present in the trajectories in the NMRlipids databank. Lipids occuring in five or less simulations (’others’) are listed in the right. d) Currently available binary mixtures in the NMRlipids databank. e) Distribution of force fields in the simulations in the NMRlipids databank. The figures and numbers are created on 9th of May 2022 with stats.ipynb.

Structure of the NMRlipids databank. As illustrated in Fig. 2., the script creates a README.yaml file that contains all the essential information of an added simulation based on the information given according to the instructions. The created README.yaml files are stored in folders in Data/Simulations. The folders are named after the hash identities of trajectory and topology files. While the raw simulation data is not directly stored in the NMRlipids databank, the README.yaml files contain permanent links from where the raw data can be accessed when needed.

For the quality evaluation, simulations are connected to the available experimental C-H bond order parameters from NMR and x-ray scattering from factors, which are also included in the NMRlipids databank. The connection between a simulation and experimental data set is made by the script when molar concentrations of all molecules are within ±5 percentage units, charged lipids have the same counterions, and temperature is within ±2 degrees. In such cases, the paths to the experimental data are added into the simulation README.yaml file.

Figure 2: Figure 2: Structure of the NMRlipids databank. Manually added input data (blue boxes) includes basic information on the simulation (more details from here), permanent links to the raw data, and experimental data if available. The databank entries (red box) and analysis results (green boxes) are automatically generated by the computer programs included in the NMRlipids databank (yellow boxes) and stored in here. Because raw data are not permanently stored in the NMRlipids databank but can be accessed based on the information in the databank, this connection is marked with the dashed line.

Analysing simulations in the NMRlipids databank. Because README.yaml files contain all the essential information from each simulation, including the permanent location of raw data and unique naming convention for all atoms and molecules (see below), arbitrary analyses of simulations can be automatically performed for all simulations in the NMRlipids databank. For example, the code that calculates all C-H bond order parameters of all systems first loops over all README.yaml files (i.e., simulations) in the NMRlipids databank, then downloads the raw simulation to a local computer if needed, and then uses the information about the atom and molecule naming conventions in README.yaml and mapping files to perform the desired analyses. A minimal example of an analysis code is available in here. Results for order parameters, form factors, area per lipid and thickness are stored in same locations as README.yaml files. Further analyses can be conventiently stored in separate repositories with the same folder structure based on hash identities of trajectory and topology files as done, for example, for the preparation of the NMRlipids databank manuscript.  

Molecule and atom naming convention. Unique naming convetions for molecules and atoms are needed for automatic analyses over large sets of simulation data in the NMRlipids databank. Because such convention was not available for lipids, we have generated mapping files (available in here) that connect lipid atoms names in each simulation to the universal atom names and universal abbreviations of lipid names for the NMRlipids databank (see the second table in here). For a new entry into the NMRlipids databank, the universal abbreviation for each lipid and the corresponding mapping file are given as input in the COMPOSITION dictionary. The numbers of each molecule in the simulation are then automatically calculated by the (see figure 2) and stored in the COMPOSITION dictionary in README.yaml files. This information enables selection of any molecule or atom when analysing simulations in the NMRlipids databank.

Wednesday, April 20, 2022

New yaml format of mapping files

Mapping files for lipid atom names were introduced in the NMRlipids project to enable flexible analyses over simulations with different force fields and atom naming conventions. In the mapping file format described in the original post, the first column defines the universal atom name based on its attachment to lipid glycerol backbone carbons and second column gives the topology dependent name in a specific simulation. This format has been highly useful in many projects, including the NMRlipids databank development.

However, also other classifications than the name might be useful for a given atom in some applications. For example, the NMRlipids quality ranking can be made separately for lipid headgroup and tails, and monitoring, for example, flip-flops of lipid headgroups from one leaflet to another might be interesting. For such applications we need information whether a specific atoms belongs to the headgroup or acyl chain region. 

To enable such analyses we have now updated the format of mapping files to yaml in the NMRlipids databank. These files can be directly read as dictionaries in Python, thereby enabling more flexibility in terms of adding new information in the mapping files. The universal atom names are defined as in the original format, but they are now given as the keys for a dictionary. The values of these keys are another dictionary containing the topology specific atom names and the position in a lipid (headgroup, tails, etc.). The mapping file for CHARMM36 POPC exemplified in the original post looks like this in the new format:

  FRAGMENT: glycerol backbone
  FRAGMENT: glycerol backbone
  FRAGMENT: glycerol backbone
  FRAGMENT: glycerol backbone
  FRAGMENT: sn-1
  FRAGMENT: sn-1



This format enables to extend the information given in the mapping files by adding new subdictionaries as values to the key defining the universal atom name.

While the original universal naming convention was designed for glycerolipids with glycerol backbone, the mapping files have been generated also for other types of molecules, such as cholesterol and dihexadecyldimethylammonium. While the mapping are applicable also for such situation, the universal atom naming convention has to be defined. For example for cholesterol, the universal atom naming convention was based on Fig. 4 in this publication.




Monday, February 7, 2022

NMRLipids VI: Status update on the simulations with AMOEBA force field

In the first part of the NMRlipids VI project on polarisable force fields, we have focused on the CHARMM-Drude polarisable force field and analyzed the POPC bilayers at various NaCl and CaCl2 concentrations as well as POPE bilayers without salt. Our results have been summarized in a previous blog posts and the manuscript in GitHub. As a quick recap, we found that CHARMM-Drude predicts a forking for the head group beta and alpha order parameters, which is not correct and not present in CHARMM36. We confirmed that this is due to the fitting of order parameters to the wrong target values. We further found that sodium binding to POPC bilayers is too strong. As the fitting-target data for the CHARMM-Drude force field is not correct, we could not get useful information on the effect of polarizability. During the latest NMRlipids meeting in September 2021 in Prague, we decided to include POPE and POPC simulations with the AMOEBA force field into the project if possible. This was mostly due to the public interest in the AMOEBA force field, and to get a more complete picture of the available polarisable models; furthermore, we wanted to see if ion binding is good in this polarizable model — as hinted by the recent developments in the electronic-continuum-correction (ECC) force fields. Here we summarize our current experience in attempting to run AMOEBA simulations.

Since the NMRlipids meeting in Prague, we have worked on obtaining the AMOEBA force field parameters, cross-checking them for errors, and setting up the same systems we had for CHARMM-Drude. After the NMRlipids Databank meeting online in December 2021, we finalized the list of simulations to run and distributed them among the NMRlipids contributors interested in running them. Our plan was to simulate pure POPE and POPC bilayers, as well as POPC bilayers at 350 mM, 450 mM, and 1000 mM NaCl and CaCl2 concentrations. We obtained the POPC force field parameters from the authors ahead of publication and created the initial structures for the aforementioned systems.

Our initial understanding was that the AMOEBA simulations would need to be run using either Tinker or the Tinker/OpenMM interface. To this end, we prepared all the files compatible with the Tinker software. As Tinker does not support semi-isotropic pressure coupling, we used the fully anisotropic coupling. However, after running the simulations for a few nanoseconds, we observed a drastic decrease in the area per lipid values, and visual inspection of the trajectories showed that the membranes transitioned into the solid gel-like phase instead of being at the liquid-disordered phase. 

To test if the ordering of the membrane originated from the parameters or from the algorithms used, we started to search alternative options to run the AMOEBA membrane simulations.  It turned out, contrary to our initial understanding, that AMOEBA simulations can indeed be run with OpenMM. For that, the Tinker force field files need to be converted into OpenMM format, which is achievable via publicly available scripts. The conversion is only possible, unfortunately, for the Allinger-type out-of-plane-bending angles and energies. Allinger type has been used in the force field for the POPE, but POPC bilayer uses Wilson–Decius–Cross  (WDC) formulation of the out-of-plane-bend angle, which is not supported by the current version of OpenMM. Therefore, the POPE bilayer simulations using AMOEBA seem possible to run with the OpenMM engine — but the POPC not. We are in contact with the OpenMM developers to see if the WDC formulation can be implemented, but we believe it is not realistic to expect a very rapid fix to this problem.

Interestingly, according to the original publication for DOPC lipids, OpenMM was used for the simulations, although WDC has never been part of OpenMM. It is not yet clear if the authors used an in-house implementation to add WDC functionality to OpenMM, or if there is simply a typo in the publication. 

In conclusion, we should be able to run the POPE simulations with the correct pressure coupling and barostat — but for POPC this does not seem possible. We have not yet been able to run AMOEBA simulations of lipid bilayers that would remain in the liquid disordered phase. We are not yet sure if this is due to the parameters we have or the simulation algorithms we can use. We are still planning to use some time to understand this, but we are also open to alternative options to focus on ion-binding affinity. For example, one option would be to run simulations in the NVT ensemble with area per lipid fixed to a reasonable value — yet such a membrane would be under tension, and the area per lipid could not equilibrate upon ion binding, which could potentially affect the results. However, such simulations could still give valuable information on how polarizability affects ion binding (while we wait for fully functional methods for polarizable membrane simulations to become available). In addition, the NVT ensemble might be able to give us useful results on the head group order parameters. In any case, until the WDC angle gets implemented in OpenMM it, unfortunately, seems that we have no way to create the exact equivalents of our CHARMM-Drude simulation setups.

In order to discuss these findings and possible ways to continue this project, we will have a Zoom discussion on 10th of March 2022 at 15:00 CET. If you are interested to join, send email (b.kav at for the Zoom link.

Batuhan Kav