Insights into the classification of small GTPases

Figures & data

Figure 1 Boxplot of descriptor performances (mean AUC). On the y-axis the AUC values for all descriptors analyzed are shown as a boxplot.

Table 1 Normalized positional residue frequency at helix termini N1. The descriptor values for each amino acid (single letter code) are shown

Amino acid	Value
A	1.10
R	1.05
N	0.72
D	1.14
C	0.26
Q	1.31
E	2.30
G	0.55
H	0.83
I	1.06
L	0.84
K	1.08
M	0.90
F	0.90
P	1.67
S	0.81
T	0.77
W	1.26
Y	0.99
V	0.76

Figure 2 Structural plot of descriptor for all families.

Notes: The x-axis represents the normalized sequence position, whereas the y-axis denotes the descriptor value. The interval of all existing values at each position is shown for each family. A: Arf; B: Rab; C: Ran; D: Rho.

Figure 3 ROC curve of the best performing random forest. (1-specificity) against sensitivity, ranging from 0 to 1 on both axes.

Figure 4 Importance plot of the GTPases classification. The x-axis represents the normalized sequence positions, whereas the y-axis denotes the percental increase in misclassification rate.

Figure 5 Most important positions for the identification of small GTPases. The most important regions (>4%) for the discrimination whether a protein belongs to the class of small GTPases are highlighted in red within the Rab6A structure.²⁷ Furthermore, the most important positions within these (>10%), are highlighted in magenta. The C-terminal part is unstructured, and thus, not shown. The Mg ion is shown as a sphere.

Table 2 Family classification. The mean AUC values, standard deviations (sd) and coefficient of variation (cv) are shown for each family of small GTPases

Family	AUC	SD	CV
Rab	0.9979	0.0001	0.0001
Rho	0.9988	0.0001	0.0001
Arf/Sar	0.9998	0.0001	0.0001
Ran	0.9999	0.0001	0.0001

Figure 6 Sequence alignment.

Notes: yellow: Rab subfamily specific motifs RabSF1–4; red: Rab family specific motifs RabF1–5; green: Phosphate/Magnesium and Guanine binding (PM/G) motifs; gray: important residues identified for family classification; hhhh: α-helices 1–5; eeee: β-sheets 1–6; swI, swII: switch 1 and 2. The alignment was created manually to accommodate structural and functional sequence elements defined by Pereira-Leal and Seabra (2000) and others.²⁸

Figure 7 Most important positions for the classification of small GTPase families.

The most important regions (>4%) for the classification of whether a protein belongs to a specific family are highlighted in red. The most important positions within the Ran subfamily is located in the C-terminal part of the protein, in an unstructured region, which is not shown here. Mg and Ca are shown as spheres.

A: Structure of Rab6A;²⁷

B: Structure of Arf1;³⁰

C: Structure of Rho6.³¹

Figure 8 Classification processing flow. A sequence is only forwarded as an input sequence to the subfamily RFs, if it was assigned and identified as a GTPase by the GTPase-RF. If the highest output value of a family of RFs exceeds 0.5 for such an input sequence, the protein sequence is assigned to this specific family.

RecachaRBouletAJollivetFStructural basis for recruitment of Rab6-interacting protein 1 to Golgi via a RUN domainStructure200917213019141279

 PubMed Web of Science ®Google Scholar

Pereira-LealJBSeabraMCThe mammalian Rab family of small GTPases: definition of family and subfamily sequence motifs suggests a mechanism for functional specificity in the Ras superfamilyJ Mol Biol20003011077108710966806

 PubMed Web of Science ®Google Scholar

LiuYKahnRAPrestegardJHStructure and membrane interaction of myristoylated ARF1Structure200917798719141284

 PubMed Web of Science ®Google Scholar

TongYTempelWShenLCrystal structure of the effector domain of PLXNB1 bound with Rnd1 GTPaseIn press10.210/pdb2rex/pdb

 Google Scholar