Supplementary information repository for the paper:

A. B. B. Lima, A. F. Pereira, L. V. Araújo, W. Tárraga, C. A. Tavares, J. de Oliveira Araújo, L. Martínez, Osmolyte Effects on Protein Conformations: Self-contained, Free, Fast Implementations of Additive Transfer Models with Novel Glycine-Activity Corrections for Multi-Cosolvent Predictions, 2026.

This package contains the validation scripts, reference datasets, and analysis code used to benchmark the Auton–Bolen (AB), Moeser–Horinek (MH), and Accessibility implementations available in PDBTools.jl.

• Validation against the AB server — m-value predictions for 36 protein structures compared against outputs of the Auton–Bolen web server, using both server SASAs and Creamer-model SASAs.
• Reproduction of the Creamer denatured-state SASA model — from-scratch computation of backbone and side-chain SASA statistics for all 20 residue types, validated against Creamer's published values for 43 reference proteins.
• MH vs AB comparison — systematic comparison of total, backbone, and side-chain m-value contributions, revealing large model differences in backbone/side-chain partitioning despite good agreement in total m-values.
• Accessibility model — The MH universal backbone treatment to all nine cosolvents covered by the AB parameterization, using apparent transfer free energies.
• Comparison with experimental m-values — benchmark of MH, AB, and Accessibility predictions against experimental urea m-values and other osmolytes, including the Rydeen et al. (2018) dataset for SH3 unfolding and GB1 dimer dissociation.
• Backbone and side-chain contributions across fold families — analysis of per-residue m-value contributions across ~15k non-homologous domains from the CATH-S20 database, classified by secondary structure content.

Julia 1.10 or greater is required. Install Julia or use juliaup to manage versions.

julia> import Pkg

julia> Pkg.develop(url="https://github.com/m3g/LAPM.jl")

julia> using LAPM

Full documentation with all validation plots is available at: https://m3g.github.io/LAPM.jl

The transfer free energy models themselves are documented in PDBTools.jl: https://m3g.github.io/PDBTools.jl

The implementations benchmarked here are part of PDBTools.jl:

L. Martínez, A. B. B. Lima, PDBTools.jl: A Lightweight and High-Performance Julia Package for Molecular Structure File Handling and Analysis, J. Chem. Inf. Model., 2026. https://doi.org/10.1021/acs.jcim.6c00847