[Update 12.11.2014] Originally this post claimed that the GAFFlipid model has the wrong isomer for the glycerol backbone. Further studies revealed that the problem was actually the initial configuration downloaded from http://lipidbook.bioch.ox.ac.uk/package/show/id/150.html having different isomers in different leaflets (see comments to this post). After setting up new simulations with correct initial stucture also the GAFFlipid has the correct stereospecifity for the g\(_1\) segment:
[Original post begins here:]
Based on discussions in the blog and the experiments by Gally et al. we have now reported the order parameter results by taking the stereospecifity of g\(_1\) and g\(_3\) into account. The result was already reported in the Current status of the project on 29.4.2014, and it is shown here again in Fig. 1.
 |
| Figure 1. Order parameters for choline and glycerol for POPC. Magnitudes measured by Ferreira et al., signs measured by Hong et al., Hong et al. and Gross et al. , and stereospecific labeling measured by Gally et al. |
[Original post begins here:]
Based on discussions in the blog and the experiments by Gally et al. we have now reported the order parameter results by taking the stereospecifity of g\(_1\) and g\(_3\) into account. The result was already reported in the Current status of the project on 29.4.2014, and it is shown here again in Fig. 1.
 |
| Figure 1. Order parameters for choline and glycerol for POPC. Magnitudes measured by Ferreira et al., signs measured by Hong et al., Hong et al. and Gross et al. , and stereospecific labeling measured by Gally et al. |
The main conclusion from the Fig. 1 was that the GAFFlipid model has different stereospecifity in g\(_1\) order parameters compared to CHARMM36, MacRog and experiments. While writing the new version of manuscript I was plotting the lipid structures from different simulations and noticed that the GAFFlipid has different isomer in glycerol compared to other models. This is demonstrated in Fig. 2
 |
| Figure 2. Snapshots of lipid structures from simulations with different models. All the structures are shown such that the viewer is looking at the direction of g\(_1\)-g\(2\) bond. In Berger, MacRog and CHARMM36 the g\(_2\) hydrogen is pointing to the right. In GAFFlipid it is pointing left. |
In conclusion, the different isomer in GAFFlipid explains the different R/S labeling observed in simulations. I did also check the isomer in Lipid14 and it seems that the issue has been fixed there (Lipid14 had the magnitudes more off from experiments, however, as seen in the The lipid forcefield comparison post).
