3.5 Evaluating
the A3D Output
The evaluation of A3D predictions and the subsequent identifica-
tion of relevant APRs would benefit from a transversal handling of
the four outputs provided by A3D. This would allow the user from
a generic landscape of amino acidic propensities to the identifica-
tion of genuine specific structural APRs. There are five crucial issues
that the user should be warned about before evaluating any A3D
run:- Maximum knowledge of the protein under study is critical to
attain a reliable interpretation of the A3D output. - The A3D threshold is equal to 0. Thus, a positive A3Dscore
implies aggregation prone residues and structures, and vice
versa. - Buried residues are considered as noninfluencing aggregation
under native conditions, therefore, they are assigned with an
A3Dscoreof 0. - Positive scores of A3D do not necessarily reflect deleterious
APR, these regions might be functional in a biological context. - A3D-plot and score-table interfaces are represented on the
amino acidic sequence basis, but they reflect structural propen-
sities (seeNote 15).
When predicting APRs over a specific structure in static or
dynamic mode, the user should first identify the A3D positive
amino acids or regions on the protein sequence. The A3D-plot
and score-table combination (Fig.3a) allow identifying rapidly
those putative regions that exceed the threshold. By taking advan-
tage of theJSmoltool, those preselected regions can be mapped on
the protein surface (red color if positive, blue if negative). A visual
inspection of the spatial vicinity—theJSmolinterface provides the
identities of selected amino acids- which would allow to find the
neighboring residues contributing to a given structural APR, since,
usually, these amino acids are not consecutive in sequence (Fig.3a
andc). The dynamic interface of A3D provides information on the
global plasticity of the protein of interest, regions with high flexi-
bility displaying higher RMSDs (seeNote 16). Thus, when an
identified APR colocalizes with regions exhibiting high RMSDs
values is indicative that this aggregation-prone cluster is being
exposed due to conformational fluctuations. On the contrary, posi-
tive residues colocalizing with stable nondynamic regions of the
protein are always likely to be aggregation-prone and exposed to
the solvent.
On the other side, when handling large amounts of data, the
user should take advantage of the global numeric parameters from
the table-score, dissected inNote 12. In particular, theaverage
scoreis the best parameter to compare different proteins, and the
total scoreis optimal to scrutinize the effect of mutations over a
defined protein.436 Jordi Pujols et al.