Back

Creating Coarse-Grained Systems with COBY: Towards Higher Accuracy in Membrane Complexity

Andreasen, M. D.; Souza, P. C. T.; Schott, B.; Zuzic, L.

2024-07-23 biophysics
10.1101/2024.07.23.604601 bioRxiv
Show abstract

Current trends in molecular modelling are geared towards increasingly realistic representations of the modelled systems. This is reflected in larger, more complex systems, which are difficult to build and would ideally rely on a software that converts userprovided descriptors into system coordinates. This is not a trivial task, as the building algorithms use simplifications that can introduce inaccuracies in the system properties that do not correspond to the requested values. We created COBY, a coarse-grained system builder that can create a large variety of systems in a single command call using Martini molecule models. We improved the accuracy of the complex membrane and solvent building procedures, introduced a variety of arguments that can be used to build diverse systems, and implemented features intended for force field development. COBY can be used to build flat membranes of any degree of complexity, handle protein and solvent insertion, solute flooding, stacked membranes, membrane patches and pores, and includes advanced functionalities such as molecule import, lipid building from fragments, handling of multiple parameter libraries, and several choices of algorithms for interpreting user-provided system descriptors. COBY is an open-source software written in Python 3, and the code, documentation, and tutorials are hosted at github.com/MikkelDA/COBY. TOC Graphic O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=110 SRC="FIGDIR/small/604601v1_ufig1.gif" ALT="Figure 1"> View larger version (39K): org.highwire.dtl.DTLVardef@15e46a4org.highwire.dtl.DTLVardef@6fd668org.highwire.dtl.DTLVardef@657541org.highwire.dtl.DTLVardef@f51840_HPS_FORMAT_FIGEXP M_FIG C_FIG

Matching journals

The top 2 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.