Background Representation Glucose Binding Hexose Binding Rules Characterizing and Predicting Hexose-Binding Sites Houssam Nassif http://pages.cs.wisc.edu/~hous21/ CIBM Seminar 25 January 2011
Background Representation Glucose Binding Hexose Binding Rules Outline Background 1 Motivation Hexoses Atomic Interactions Problem Representation 2 Glucose Binding Modeling 3 Classification Approach Results Hexose Binding Rules Empirical Generation 4 Rule Inference Results
Background Representation Glucose Binding Hexose Binding Rules Outline Background 1 Motivation Hexoses Atomic Interactions Problem Representation 2 Glucose Binding Modeling 3 Classification Approach Results Hexose Binding Rules Empirical Generation 4 Rule Inference Results
Background Representation Glucose Binding Hexose Binding Rules Hexoses Pathways 6-carbon sugar molecules Key role in several biochemical pathways cellular energy release signaling carbohydrate synthesis regulation of gene expression . . .
Background Representation Glucose Binding Hexose Binding Rules Tasks Galactose, glucose, mannose High specificity to diverse protein families Lack of glucose model No data-driven comparison to biochemical findings Tasks Glucose-binding model Empirical comparison to wet-lab findings
Background Representation Glucose Binding Hexose Binding Rules Tasks Galactose, glucose, mannose High specificity to diverse protein families Lack of glucose model No data-driven comparison to biochemical findings Tasks Glucose-binding model Empirical comparison to wet-lab findings
Background Representation Glucose Binding Hexose Binding Rules Outline Background 1 Motivation Hexoses Atomic Interactions Problem Representation 2 Glucose Binding Modeling 3 Classification Approach Results Hexose Binding Rules Empirical Generation 4 Rule Inference Results
Background Representation Glucose Binding Hexose Binding Rules Hexose Stereoisomers H ✧ O H ✧ O H ✧ O ❜ ✧ ❜ ✧ ❜ ✧ ❜ ✧ ❜ ✧ ❜ ✧ ✧ ✧ ✧ C C C H C OH H C OH OH C H OH C H OH C H OH C H OH C H H C OH H C OH H C OH H C OH H C OH CH 2 OH CH 2 OH CH 2 OH (a) (b) (c) Figure: (a) D -Galactose; (b) D -Glucose; (c) D -Mannose
Background Representation Glucose Binding Hexose Binding Rules Hexose Structure H ✧ O ❜ ✧ ❜ ✧ ✧ C Contains two functional groups H C OH Both groups can interact together OH C H H C OH ❚ ❚ H C OH C O ✔ OH ✔ CH 2 OH Figure: Glucose (a) Carbonyl (b) Hydroxyl
Background Representation Glucose Binding Hexose Binding Rules Hexose Cyclization The molecule folds on itself and forms a pyranose ring. In two different ways. Watch the star! H ✧ O ❜ ✧ ❜ ✧ CH 2 OH ✧ CH 2 OH C O H O OH* H H ✔ ✔ H C OH ❚ ❚ H H ✔ ❚ ✔ ❚ OH H OH H ❚ ✔ ❚ ✔ ⇀ ⇀ OH C H − − − − ↽ ↽ − − − − OH OH* OH H ❚ ✔ ❚ ✔ H C OH H OH H OH H C OH (a) α -pyranose (c) β -pyranose CH 2 OH (b) Open chain
Background Representation Glucose Binding Hexose Binding Rules Outline Background 1 Motivation Hexoses Atomic Interactions Problem Representation 2 Glucose Binding Modeling 3 Classification Approach Results Hexose Binding Rules Empirical Generation 4 Rule Inference Results
Background Representation Glucose Binding Hexose Binding Rules Covalent Bonds Close and strong interaction Forms a molecule O − δ H + δ ✔ ✔ Atoms share electrons Electronegativity: Equal ⇒ nonpolar Figure: Covalent Different ⇒ polar bond Partial charges Definition Electronegativity: Measure of atom’s attraction for electrons
Background Representation Glucose Binding Hexose Binding Rules Covalent Bonds Close and strong interaction Forms a molecule O − δ H + δ ✔ ✔ Atoms share electrons Electronegativity: Equal ⇒ nonpolar Figure: Covalent Different ⇒ polar bond Partial charges Definition Electronegativity: Measure of atom’s attraction for electrons
Background Representation Glucose Binding Hexose Binding Rules Covalent Bonds Close and strong interaction Forms a molecule O − δ H + δ ✔ ✔ Atoms share electrons Electronegativity: Equal ⇒ nonpolar Figure: Covalent Different ⇒ polar polar bond Partial charges Definition Electronegativity: Measure of atom’s attraction for electrons
Background Representation Glucose Binding Hexose Binding Rules Hydrogen Bonds H Attraction between a positively + δ - δ charged H and a negatively O H N H charged atom ✔ ✔ H H Hexose attaches to the protein using hydrogen bonds Figure: Hydrogen bond
Background Representation Glucose Binding Hexose Binding Rules Van der Waals and Hydrophobicity Definition Van der Waals Forces: Weak electrostatic attraction and repulsion forces Definition ( Hydrophobicity ) Hydrophobic: water hating. Hydrophilic: water loving. Hydrophobic/Hydrophilic atoms tend to gather together. Dual nature: Pyranose ring is hydrophobic Hydroxyl group is hydrophilic
Background Representation Glucose Binding Hexose Binding Rules Van der Waals and Hydrophobicity Definition Van der Waals Forces: Weak electrostatic attraction and repulsion forces Definition ( Hydrophobicity ) Hydrophobic: water hating. Hydrophilic: water loving. Hydrophobic/Hydrophilic atoms tend to gather together. Dual nature: Pyranose ring is hydrophobic Hydroxyl group is hydrophilic
Background Representation Glucose Binding Hexose Binding Rules Outline Background 1 Motivation Hexoses Atomic Interactions Problem Representation 2 Glucose Binding Modeling 3 Classification Approach Results Hexose Binding Rules Empirical Generation 4 Rule Inference Results
Background Representation Glucose Binding Hexose Binding Rules Binding-Site Representation
Background Representation Glucose Binding Hexose Binding Rules Binding-Site Feature Extraction 1: procedure E XTRACT F EATURES (binding site center) for all concentric layers do 2: for all PDB atoms do 3: get coordinates 4: get charge 5: get hydrophobicity 6: get hydrogen-bonding 7: get residue 8: 9: end for 10: end for 11: end procedure
Background Representation Glucose Binding Hexose Binding Rules Binding-Site Features Atomic Feature Values Charge Negative, Neutral, Positive Hydrogen-bonding Non-hydrogen bonding, Hydrogen-bonding Hydrophobicity Hydrophilic, Hydroneutral, Hydrophobic Residue Grouping Amino Acids Aromatic H IS , P HE , T RP , T YR Aliphatic A LA , I LE , L EU , M ET , V AL Neutral A SN , C YS , G LN , G LY , P RO , S ER , T HR Acidic A SP , G LU Basic A RG , L YS
Background Representation Glucose Binding Hexose Binding Rules Data Mining Empirical evidence suggests that hexose docking is not accompanied by protein conformational changes (galactose) Hexose dataset Mine PDB for glucose/hexoses Discard theoretical structures and redundancies Discard covalently bound and floating in medium Impose 30 % cut-off overall sequence identity Discard if other ligands bind or are present Non-hexose dataset Non-sugar binding sites Glucose/hexose-like binding sites Random non-binding sites
Background Representation Glucose Binding Hexose Binding Rules Classifier Outline a) Known glucose binding d) Site feature g) sites vector Classifier (training phase) b) Known non-glucose e) Non-Site sites feature vector i) Glucose binding site h) classifier c) Unknown site f) Unknown site (testing feature vector j) Not a glucose phase) binding site
Background Representation Glucose Binding Hexose Binding Rules Outline Background 1 Motivation Hexoses Atomic Interactions Problem Representation 2 Glucose Binding Modeling 3 Classification Approach Results Hexose Binding Rules Empirical Generation 4 Rule Inference Results
Background Representation Glucose Binding Hexose Binding Rules Support Vector Machines (SVM) Construct the optimal separating hyperplane (usually in a higher feature space) Maximize margins : minimal distance from the hyperplane Only Support Vectors (SV) specify the margins/hyperplane Small number of SV ⇔ good generalization
Background Representation Glucose Binding Hexose Binding Rules Support Vector Machines (SVM) Construct the optimal separating hyperplane (usually in a higher feature space) Maximize margins : minimal distance from the hyperplane Only Support Vectors (SV) specify the margins/hyperplane Small number of SV ⇔ good generalization
Background Representation Glucose Binding Hexose Binding Rules Support Vector Machines (SVM) Construct the optimal separating hyperplane (usually in a higher feature space) Maximize margins : minimal distance from the hyperplane Only Support Vectors (SV) specify the margins/hyperplane Small number of SV ⇔ good generalization
Background Representation Glucose Binding Hexose Binding Rules Support Vector Machines (SVM) Construct the optimal separating hyperplane (usually in a higher feature space) Maximize margins : minimal distance from the hyperplane Only Support Vectors (SV) specify the margins/hyperplane Small number of SV ⇔ good generalization
Recommend
More recommend
