Prediction of TF target sites based on atomistic models of protein-DNA complexes

BMC Bioinformatics

Table 3 Comparison of cumulative contact and readout position weight matrices for 4 prokaryotic (top) and 4 eukaryotic (bottom) transcription factors

SCOP v1.73 superfamily	TF [PDB id]	Resolution (Å)	R_obs	E-value_contacts	E-value_readout
Winged helix	CRP [1cgp]	3	0.24	7.93E-03	3.76E-05
C-terminal domain of the bipartite response regulators	NarL [1je8]	2.12	0.23	3.58E-05	7.01E-07
lambda repressor-like DNA-binding domains	PurR [2pua]	2.9	0.16	4.33E-15	5.51E-01 (7.58E-04)
ribbon-helix-helix	MetJ [1cma]	2.8	0.22	1.22E-01	3.30E-01
Zn2/Cys6 DNA-binding domain	LEU3 [2er8]	2.85	0.23	4.97E-06	5.52E-05
HLH, helix-loop-helix DNA-binding domain	PHO4 [1a0a]	2.8	0.23	3.57E-07	3.97E-07
Homeodomain-like	RAP1 [1ign]	2.25	0.21	5.52E-03	1.89E-02 (6.40E-04)
C2H2 and C2HC zinc fingers	ZIF268 [1aay]	1.6	0.19	7.93E-14	1.99E-14

In the first column we include the name of the structural superfamily of each TF according to SCOP [37]. In the second column the name of each TF is followed by the PDB identifier of the structure used enclosed in brackets. The following two columns contain the resolution and the R-value of the crystals respectively. The last two columns report the E-values resulting from STAMP [38] local alignments between the structure-based PWMs obtained with the method of Morozov and Siggia [32] (E-value_contacts) or our approach (E-value_readout) and the corresponding cognate PWMs, extracted from the literature [43, 49, 53]. Parenthesized E-values were calculated after sampling interface rotamers with DNAPROT.

ISSN: 1471-2105