Analysis of the functional regions predicted by PatchFinder
PatchFinder detects regions of evolutionarily conserved residues on the
protein surface in proteins with known 3D structure (Nimrod et al., 2005; Nimrod et
al., 2008). These regions (patches) typically correspond to the functional regions of
these proteins. We examined PatchFinder predictions on a dataset of 121 DBPs which
were co-crystallized with DNA. In 74% of the cases at least half of the positions in
the patch were in contact with the DNA, but only in 17% of the cases the patch
covered at least half of that interface (Nimrod et al., 2009). We further analyzed the
protein-DNA interfaces in these structures using NUCPLOT (Luscombe et al., 1997).
According to this analysis, PatchFinder’s patches covered at least half of the interface
hydrogen bonds in 50% of the proteins (Nimrod et al., 2009).
Hydrogen bonds are often important in protein-DNA sequence recognition
(Mandel-Gutfreund et al., 1995). Distinction between protein residues that contact
the DNA bases and the rest of the interface residues (i.e. those that contact the sugarphosphate backbone) may provide some information on the residue positions that are
responsible for binding specificity (Lichtarge et al., 1997). Supplementary Figure 1
shows that in 47% of the cases at least half of the residues that contact the DNA-bases
were in the PatchFinder patch, which is more than twice the fraction that met the same
criteria for the whole interface.
Supplementary Figure 1 suggested a partitioning of PatchFinder’s patches into
two groups, according to the fraction of protein-DNA bases interactions identified by
each patch. We used the DAVID Bioinformatic Resources (Huang da et al., 2009) in
order to characterize the proteins for which PatchFinder identified the highest fraction
of protein-DNA bases contacts in comparison with the whole dataset of DBPs. The
DAVID analysis found that this group is enriched with the molecular function GO
term (Ashburner et al., 2000) ‘transcription regulator activity’ (p-value 4.1e-4 ). This
GO term is mainly associated with transcription factors which typically require
specific recognition of the DNA. By contrast, inspection of the proteins whose MLpatches were devoid of protein-DNA bases interactions showed that this group is
enriched with the GO term ‘calatlytic activity’ (p-value 7.3e-3). In these proteins, the
main conservation signal came from the catalytic site rather than the recognition site.
A similar trend was noticed with the analysis of protein-DNA interface hydrogen
bonds (Nimrod et al., 2009).
Example of a protein predicted to bind DNA
PDB entry 3BHW corresponds to the structure of an uncharacterized protein of 199
residues from Magnetospirillum magneticum, solved by the New York SGX Research
Center for Structural Genomics (NYSGXRC). The protein was assigned the PFAM
domain DUF1819 of uncharacterized proteins (Finn et al., 2008). Some of the entries
associated with this PFAM domain are putative inner membrane proteins. The iDBPs
server predicts that the protein binds DNA. We examined the protein with the Dali
server for the identification of structural similarity between proteins (Holm and Park,
2000). Dali found structural similarity between 3BHW and several DNA binding
proteins including a restriction endonuclease BpuJI from Bacillus pumilus that was
crystallized bound to a dsDNA (PDB entry 2VLA; (Sukackaite et al., 2008)).
Supplementary Figure 2.A presents a superimposition of PDB entry 3BHW (green
ribbons) and BpuJI (blue ribbons) bound to a dsDNA (orange). Besides the high
structural similarity between the proteins, the superimposition also placed 3BHW in a
potential DNA binding position with only minor clashes with the DNA. Electrostatic
potential analysis of 3BHW (Supplementary Figure 2.B), showed that the suggested
DNA binding region is positively charged as well. This property is typical of DNA
binding regions (Stawiski et al., 2003). We also examined the protein using the DBDhunter web server for the identification of DBPs. DBD-hunter is a knowledge-based
method that looks for structural similarity between the query protein and known DBPs
and analyses the DNA-protein statistical potential (Gao and Skolnick, 2008). This
approach, which is complementary to ours, also found 3BHW as a potential DBP.
Our analysis therefore provides useful information regarding the possible functional
role of this hypothetical protein, awaiting experimental validation. Moreover, we
suggest potential residues that directly participate in the predicted binding of DNA
(see Supplementary Figure 3). These positions are candidates for future mutagenesis
efforts, aimed at validating the computational functional prediction.
Supplementary Figure 1.
Analysis of PatchFinder’s predictions for DNA binding regions. The plot presents
the fraction of proteins within each sensitivity bin in a dataset of 121 DBPs which
were crystallized bound to DNA. Black bars represent the sensitivity in the
identification of protein residues which are within 6Å from the DNA (Nimrod et al.,
2009). Grey bars represent the identification of protein residues which contact DNA
bases. The calculations were conducted using NUCPLOT (Luscombe et al., 1997).
According to the analysis most of the patches are associated with one of the extreme
groups in the graph. They either cover more than 90% of the positions that contact the
DNA-bases or less than 10% of these positions. The first group is enriched with
transcription factors which typically require DNA sequence recognition. The second
group, on the other hand, is enriched with DBPs with catalytic activity.
Supplementary Figure 2.
A. Superimposition of PDB entry 3BHW from Magnetospirillum magneticum (green
ribbons) with restriction endonuclease BpuJI from Bacillus pumilus (blue ribbons)
bound to a dsDNA (orange; (Sukackaite et al., 2008)). The superimposition was
calculated using the C-alpha Match algorithm (Bachar et al., 1993) and the figure was
produced using UCSF Chimera (Pettersen et al., 2004). B. The electrostatic potential
of 3BHW mapped on the protein surface according to the coloring scale on the left.
The potentials were calculated using PMV (Sanner, 1999) and the figure was
produced with PyMol (DeLano, 2002).
Overall, the figure shows evidence that
supports the prediction that 3BHW may bind DNA.
Supplementary Figure 3.
A screenshot of PatchFinder’s analysis for PDB entry 3BHW provided by the iDBPs
server. PatchFinder found two patches with high confidence to be functionally
important. The molecule is in space-filling representation. The primary and the
secondary patches are colored red and blue respectively and the rest of the protein is
in grey. Assuming 3BHW is a DBP, these patches of evolutionarily conserved
residues are good candidates for DNA binding residues.
Similar analysis is provided by the iDBPs server for every submitted protein structure,
when enough sequence homologs are available (see main text).
References
Ashburner, M. et al. (2000) Gene ontology: tool for the unification of biology. The
Gene Ontology Consortium. Nat. Genet., 25, 25-29.
Bachar, O. et al. (1993) A computer vision based technique for 3-D sequenceindependent structural comparison of proteins. Protein Eng., 6, 279-288.
DeLano, W.L. (2002) The PyMOL molecular graphics system.
Finn, R.D. et al. (2008) The Pfam protein families database. Nucleic Acids Res, 36,
D281-288. Epub 2007 Nov 2026.
Gao, M. and Skolnick, J. (2008) DBD-Hunter: a knowledge-based method for the
prediction of DNA-protein interactions. Nucleic Acids Res, 36, 3978-3992.
Holm, L. and Park, J. (2000) DaliLite workbench for protein structure comparison.
Bioinformatics, 16, 566-567.
Huang da, W. et al. (2009) Systematic and integrative analysis of large gene lists
using DAVID bioinformatics resources. Nat Protoc, 4, 44-57.
Lichtarge, O. et al. (1997) Identification of functional surfaces of the zinc binding
domains of intracellular receptors. J Mol Biol, 274, 325-337.
Luscombe, N.M. et al. (1997) NUCPLOT: a program to generate schematic diagrams
of protein-nucleic acid interactions. Nucleic Acids Res., 25, 4940-4945.
Mandel-Gutfreund, Y. et al. (1995) Comprehensive analysis of hydrogen bonds in
regulatory protein DNA-complexes: in search of common principles. J. Mol.
Biol., 253, 370-382.
Nimrod, G. et al. (2005) In silico identification of functional regions in proteins.
Bioinformatics, 21, i328-337.
Nimrod, G. et al. (2008) Detection of functionally important regions in 'hypothetical
proteins' of known structure. Structure, 16, 1755-1763.
Nimrod, G. et al. (2009) Identification of DNA-binding proteins using structural,
electrostatic and evolutionary features. J. Mol. Biol., 387, 1040-1053.
Pettersen, E.F. et al. (2004) UCSF Chimera - a visualization system for exploratory
research and analysis. J. Comput. Chem., 25, 1605-1612.
Sanner, M.F. (1999) Python: a programming language for software integration and
development. J. Mol. Graph. Model., 17, 57-61.
Stawiski, E.W. et al. (2003) Annotating nucleic acid-binding function based on
protein structure. J. Mol. Biol., 326, 1065-1079.
Sukackaite, R. et al. (2008) The recognition domain of the BpuJI restriction
endonuclease in complex with cognate DNA at 1.3-A resolution. J Mol Biol,
378, 1084-1093.