Specificity Quantification of Biomolecular Recognition and Its Implication for Drug Discovery

Scientific Report

Specificity quantification of biomolecular recognition and its implication for drug discovery

Zhiqiang Yan, Jin Wang

Supplementary Figure 1 / Flowchart of SPA development.
Supplementary Figure 2 / Pearson correlation between predicted affinity and experimental affinity for the test set.
Supplementary Figure 3 / Evolution and convergence of Monte Carlo search with simulated annealing for the optimization of SPA.
Supplementary Table 1 / Success rates of identifying the ``native'' conformation among the top-Scored binding poses under RMSD cutoff 2.0Å.
Supplementary Table 2 / Enrichments of drugs at top 5% and 10% of total ranked ligands.
Supplementary Table 3 / 22 Atom types used for calculating interactions based on the SYBYL definition.
Supplementary Table 4 / 37 selective drugs and 20 non-selective drugs of COX-2.

Supplementary Figure 1

The development of SPA contains three stages: The design of database and development strategy, optimization of the scoring function, application of SPA.

Supplementary Figure 2

Pearson correlation between predicted affinity and experimental affinity for the test set. The correlation coefficient is 0.668, the predicted affinityis obtained by scaling the binding scores with a linear equation: y=-4.95+0.0473x, which is a fitting from the training set; the experimental affinityis estimated from Kd or Ki which are experimentally determined dissociation constant and inhibition constant respectively.

Supplementary Figure 3

Evolution and convergence of Monte Carlo search with simulated annealing for the optimization of SPA. a. Evolution of the coupled parameter ρ(=λγ) of specificity and affinity, showing the optimization process of SPA. b. Evolution of λ(red) and γ (blue), showing the maximization processes of specificity (λ) and affinity prediction (γ).

Supplementary Table 1

Success rates of identifying the ``native'' conformation among the top-scored binding poses under RMSD cutoff 2.0Å.

Scoring Functiona / One / Two / Three / Four / Five
SPA / 84.7 / 90.4 / 91.0 / 92.1 / 93.8
GOLD/ASP / 82.5 / 90.2 / 92.3 / 94.0 / 95.6
DrugScorePDB/PairSurf / 74.3 / 89.1 / 91.8 / 93.4 / 95.1
DS/PLP1 / 75.4 / 86.9 / 90.2 / 95.1 / 97.3
DS/LigScore2 / 71.6 / 85.8 / 88.0 / 92.9 / 92.9
GlideScore/SP / 73.2 / 83.1 / 86.9 / 90.2 / 93.4
Gold/GoldScore / 68.9 / 79.8 / 85.2 / 87.4 / 89.6
X-Score1.2/HMScore / 68.3 / 82.0 / 84.2 / 88.5 / 90.7
GOLD/ChemScore / 70.5 / 78.7 / 82.0 / 85.2 / 86.9
SYBYL/F-Score / 64.5 / 74.3 / 82.0 / 87.4 / 90.7
SYBYL/ChemScore / 60.1 / 72.1 / 78.7 / 83.1 / 84.7
DS/Ludi2 / 57.4 / 70.5 / 75.4 / 80.3 / 83.6
DS/Jain / 44.8 / 63.9 / 70.5 / 76.0 / 79.2
SYBYL/G-Score / 41.5 / 55.7 / 67.2 / 74.9 / 78.7
SYBYL/PMF-Score / 48.1 / 61.7 / 63.4 / 66.7 / 71.0
SYBYL/D-Score / 30.6 / 50.3 / 59.0 / 67.8 / 73.8
DS/PMF / 43.7 / 53.0 / 56.8 / 63.9 / 67.2

aThe results except SPA are obtained from the literature [24].

Supplementary Table 2

Enrichments of drugs at top 5% and 10% of total ranked ligands. The combined parameter (E & ISR) is obtained through logistic regression; the enrichment value means the percent of the drugs selected from all top ranked ligands according to the parameters.

Parameter / Selective / Non-selective
5% / 10% / 5% / 10%
Affinity (E) / 43.2 / 51.4 / 0.0 / 5.0
Specificity (ISR) / 48.7 / 56.8 / 10.0 / 25.0
E & ISR / 51.4 / 59.5 / 10.0 / 20.0

Supplementary Table 3

22 Atom types used for calculating interactions based on the SYBYL definition. The atom types can be converted from PDB files by software BABEL.

Type / Description / Type / Description
C.3 / carbon sp3 / N.4 / quaternary nitrogen
C.2 / carbon sp2 / O.3 / oxygen sp3
C.1 / carbon sp / O.2 / oxygen sp2
C.ar / carbon aromatic / O.co2 / carboxyl oxygen
C.cat / other carbons / S.3 / sulfur sp3
N.3 / nitrogen sp3 / S.2 / sulfur sp2
N.2 / nitrogen sp2 / S.O2 / sulfone sulfur
N.1 / nitrogen sp / P.3 / phosphorous sp3
N.ar / nitrogen aromatic / F / fluorine
N.p13 / trigonal nitrogen / Cl / chlorine
N.am / nitrogen amide / I/Bra / iodine/bromine

aThe atom types I and Br are grouped as one due to their low occurrences.

Supplementary Table 4

37 selective and 20 non-selective nonsteroidal anti-inflammatory drugs (NSAIDs) of COX-2. Selective drugs (red) are specific to inhibit COX-2, while non-selective drugs (blue) inhibit both the COX-2 and its isoenzyme COX-1.

Drugs / Drugs / Drugs
ns-398 / deracoxib / sulindac-sulfide
l-745337 / drf-4367 / ketoprofen
celecoxib / fr-188582 / ketorolac
rofecoxib / l-758115 / ibuprofen
dup-697 / l-768277 / flurbiprofen
jte-522 / l-778736 / tenoxicam
valdecoxib / l-784506 / piroxicam
etoricoxib / l-804600 / bromfenac
meloxicam / l761066 / carprofen
etodolac / parecoxib / droxicam
l-776967 / pd-138387 / fenoprofen
flosulide / pmi-001 / indoprofen
Sc-58125 / rs57067 / loxoprofen
644784 / sc299 / meclofenamic-acid
bms-347070 / sc558 / oxaprozin
cimicoxib / sc57666 / salicin
cis-stilbenes / svt-2016 / tiaprofenic-acid
ct-3 / t-614 / zomepirac
darbufelone / tolmetin / indomethacin