Relative accessible surface area or relative solvent accessibility (RSA) of a protein residue is a measure of residue solvent exposure. It can be calculated by formula:

${\displaystyle {\text{RSA}}={\text{ASA}}/{\text{MaxASA}}}$ ^[1]

where ASA is the solvent accessible surface area and MaxASA is the maximum possible solvent accessible surface area for the residue.^[1] Both ASA and MaxASA are commonly measured in ${\displaystyle {\mathrm {\AA} }^{2}}$.

To measure the relative solvent accessibility of the residue side-chain only, one usually takes MaxASA values that have been obtained from Gly-X-Gly tripeptides, where X is the residue of interest. Several MaxASA scales have been published^[1][2][3] and are commonly used (see Table).

In this table, the more recently published MaxASA values (from Tien et al. 2013^[1]) are systematically larger than the older values (from Miller et al. 1987^[2] or Rose et al. 1985^[3]). This discrepancy can be traced back to the conformation in which the Gly-X-Gly tripeptides are evaluated to calculate MaxASA. The earlier works used the extended conformation, with backbone angles of ${\displaystyle \phi =-120^{\circ }}$ and ${\displaystyle \psi =140^{\circ }}$.^[2][3] However, Tien et al. 2013^[1] demonstrated that tripeptides in extended conformation fall among the least-exposed conformations. The largest ASA values are consistently observed in alpha helices, with backbone angles around ${\displaystyle \phi =-50^{\circ }}$ and ${\displaystyle \psi =-45^{\circ }}$. Tien et al. 2013 recommend to use their theoretical MaxASA values (2nd column in Table), as they were obtained from a systematic enumeration of all possible conformations and likely represent a true upper bound to observable ASA.^[1]

ASA and hence RSA values are generally calculated from a protein structure, for example with the software DSSP.^[4] However, there is also an extensive literature attempting to predict RSA values from sequence data, using machine-learning approaches.^{[5] [6]}

Prediction tools

Experimentally predicting RSA is an expensive and time-consuming task. In recent decades, several computational methods have been introduced for RSA prediction.^[7][8][9]

References