Supplements
Yamanishi,Y., Pauwels, E., and Kotera,M.,
Drug side-effect prediction based on the integration of chemical and biological space
Supplemental figures and tables cited in the main manuscript
Comparison of eight different fingerprints
- Comparison of PR curves in KR chem
- Comparison of individual AUPR scores in KR chem
- Comparison of PR curves in KR Kchem+Kbio
- Comparison of individual AUPR scores in KR Kchem+Kbio
Fingerprint 1: E-State fingerprint (EStateFingerprinter) [Hall and Kier, 1995]
Fingerprint 2: CDK extended fingerprint (ExtendedFingerprinter) [Steinbeck et al.,2003]
Fingerprint 3: CDK fingerprint (Fingerprinter) [Steinbeck et al.,2003]
Fingerprint 4: Klekota-Roth fingerprint (FingerprinterTool) [Klekota and Roth, 2008]
Fingerprint 5: CDK graph only fingerprint (GraphOnlyFingerprinter) [Steinbeck et al.,2003]
Fingerprint 6: CDK hybridization fingerprint (HybridizationFingerprinter) [Steinbeck et al.,2003]
Fingerprint 7: MACCS Fingerprint (MACCSFingerprinter)
Fingerprint 8: Pubchem Fingerprint (PubchemFingerprinter)
Note that we used the Chemistry Development Kit (CDK) version 1.4.9 [Steinbeck et al., 2003] to calculate the above 8 fingerprints.
More detail information about each fingerprint can be found in the following website:
http://pele.farmbio.uu.se/nightly-1.3.1/cdk-javadoc-1.4.0/org/openscience/cdk/fingerprint/package-summary.html
Reference:
- Christoph Steinbeck, Yongquan Han, Stefan Kuhn, Oliver Horlacher, Edgar Luttmann, and Egon Willighagen. The Chemistry Development Kit (CDK): An Open-Source Java Library for Chemo- and Bioinformatics. J. Chem. Inf. Comput. Sci., 2003, 43 (2), pp 493-500.
- Hall, L.H. and Kier, L.B. , Electrotopological State Indices for Atom Types: A Novel Combination of Electronic, Topological, and Valence State Information, Journal of Chemical Information and Computer Science, 1995,35:1039-1045
- Klekota, Justin and Roth, Frederick P., Chemical substructures that enrich for biological activity, Bioinformatics, 2008, 24:2518-2525
Comparison of four different kernel functions
- Comparison of AUPR scores in KR chem
- Comparison of AUPR scores in KR Kchem+Kbio
Kernel 1: Tanimoto
Kernel 2: Linear
Kernel 3: Polynomial
Kernel 4: Gaussian RBF
Fingerprint 1: E-State fingerprint
Fingerprint 2: CDK extended fingerprint
Fingerprint 3: CDK fingerprint
Fingerprint 4: Klekota-Roth fingerprint
Fingerprint 5: CDK graph only fingerprint
Fingerprint 6: CDK hybridization fingerprint
Fingerprint 7: MACCS Fingerprint
Fingerprint 8: Pubchem Fingerprint
Note that the Tanimoto kernel matrix of Fingerprint 1 was singular and the algorithm did not work, so the corresponding result was not obtained.
Comparison with Sparse CCA
Predicted side-effects by the kernel regression method (KR) based on chemical and biological profiles for unchatecterized drugs in DrugBank
Predicted side-effects by the multiple kernel regression (MKR) method based on chemical and biological profiles for unchatecterized drugs in DrugBank
Predicted side-effects by the kernel regression (KR) method based on chemical profiles for unchatecterized drugs in DrugBank
Predicted side-effects by the kernel regression (KR) method based on biological profiles for unchatecterized drugs in DrugBank
Summary of all the prediction results for unchatecterized drugs in DrugBank
- Summarized prediction result
1st column: pubchem compound ID
2nd column: degree
3rd column: predicted side-effect term
4th column: frequency of the predicted side-effect in the reference set
5th column: prediction score based on chemical and biological profiles
6th column: prediction score based on chemical profiles
7th column: prediction score based on biological profiles
8th column: rank of the prediction score based on chemical and biological profiles
9th column: rank of the prediction score based on chemical profiles
10th column: rank of the prediction score based on biological profiles