How Do Metabolites Differ from Their Parent Molecules and How Are They Excreted?

Transcription

1 How Do Metabolites Differ from Their Parent Molecules and How Are They Excreted? Johannes Kirchmair 1, Andrew Howlett 1, Julio E. Peironcely 2,3,4, Daniel S. Murrell 1, Mark J. Williamson 1, Samuel E. Adams 1, Thomas Hankemeier 3,4, Leo van Buren 5, Guus Duchateau 5, Werner Klaffke 5 and Robert C. Glen 1 * 1 Unilever Centre for Molecular Sciences Informatics, Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge, CB2 1EW, United Kingdom 2 TNO Research Group Quality & Safety, P.O. Box 360, 3700 AJ Zeist, The Netherlands 3 Leiden/Amsterdam Center for Drug Research, Leiden University, Einsteinweg 55, 2333 CC Leiden, The Netherlands 4 Netherlands Metabolomics Centre, Einsteinweg 55, 2333 CL Leiden, The Netherlands 5 Unilever Research & Development, Olivier van Noortlaan, 3133 AT Vlaardingen, The Netherlands Corresponding Author * rcg28@cam.ac.uk, phone +44 (1223)

2 MATERIALS AND METHODS Hardware and Operating System All MOE and Maestro calculations were processed on a computer equipped with an Intel Core 2 Duo E GHz CPU and 8 GB of RAM running Ubuntu 9.10 Linux. Pipeline Pilot Student Edition and InChI were run on the identical machine under Windows 7. MetaPrint2D, any other Java programs and R for data analysis (RStudio version ) were run on an Apple MacBook Air with an Intel Core i7 1.8GHz CPU and 4 GB of RAM under Mac OS X version Flow diagrams were created using Aris Express version Additional Information to Question 1. Data sources and preparation. 92 molecules of the HMDB are known drug molecules 3 and were removed from the dataset. Data from the TCM Database@Taiwan are provided as individual mol2 files. These were converted into SD file format using Schrödinger Maestro Suite version Molecules from all databases were required to contain at least one carbon, nitrogen, oxygen, sulfur or phosphorous atom. Descriptor Calculation Using MOE. All molecules were imported to MOE. A few molecules with structural errors (e.g. negatively charged carbons) were identified in the converted TCM Database@Taiwan and were omitted. All molecules were processed using the Wash molecule function (default settings: explicit hydrogen atoms added, strong acids and bases deprotonated, partial charges added after hydrogen atoms and lone pairs adjusted as required for the use of MMFF94). Defined in the MOE Wash procedure, functional groups are only classified as acidic or basic if the pk a of the functional group is not close to 7. Three-dimensional representations were calculated using the Rebuild 3D function (chirality preserved, default settings). These structures were minimized using the Energy Minimize function (default settings), followed by the calculation of chemical descriptors. The MOE a_acid descriptor was used to quantify the number of acidic atoms by considering the number of acidic atoms present in acidic functional groups: e.g. a carboxylic acid consists of two acidic atoms; a phosphate group contains three acidic atoms. Analogous to this, the a_base descriptor was used to quantify the number of basic atoms present in basic functional groups. LogP was calculated using the MOE logp octanol/water model. For the correct calculation of this descriptor and descriptors counting hydrogen bond acceptors and donors neutralized representations of the chemical structures were used. Clustering for Diversity Parameter Setup. Extended connectivity fingerprints (ECFP4) were used as descriptors in combination with an RMSD distance function for clustering with Pipeline Pilot Student Edition version The objective was to obtain clusters comprised of molecules of a single molecular scaffold. The maximum distance between a cluster member and the cluster center was set to 0.5 and the average number of molecules per cluster was set to 3. These parameters were derived from iterative visual inspection of the clusters. 2

3 Additional Information to Question 2. Data sources and preparation. In addition to single-step reaction entries (a substrate transformed into a metabolite by a single enzymatic reaction) the AMDB also contains multi-step reaction entries, which condense the structural changes caused by more than one metabolic reaction into a single database entry (containing the substrate and a terminal metabolite). These entries were excluded from our analysis for redundancy reasons, which leaves biotransformations ( molecules) in total for further processing. An in-house Java application was used to remove invalid molecules (i.e. molecules with incorrect valence formulas, molecules that include an R-group notation or entries with the molecular structure missing), resulting in a total of molecules for further processing. Removal of invalid molecules leads to incomplete metabolic schemes in certain cases (i.e. schemes that do not contain at least a single top-level substrate and a single terminal metabolite). Filtering these incomplete entries resulted in a total of molecules (14272 metabolic schemes, with their substrates, intermediates and terminal metabolites). Assignment of Chemical Spaces. InChI codes were calculated for all structures using InChI for Windows version Stereochemical information was discarded for the reason that most of the investigated physicochemical property descriptors and the MetaPrint2D reaction classification are not sensitive to chirality. This also avoids missing molecules present in different datasets due to inaccurate or incomplete stereochemical characterization. Mobile hydrogen perception was enabled. Advanced tautomerism detection was activated to allow the identification of keto-enol tautomers. Additional Information to Question 3. Data sources and preparation. Biotransformations were not considered if they (i) miss the molecular structure of the substrate or metabolite or (ii) MetaPrint2D reports structural changes as being a result of more than a single enzymatic transformation or (iii) MetaPrint2D reports the reaction type as not determinable/unknown. This is to make sure that the biotransformations are based on a single enzymatic reaction. The processing resulted in reactions in total. 3

4 REFERENCES 1. RStudio, ; RStudio Project: (accessed Dec 11, 2012). 2. Aris Express, 2.3; Software AG: Darmstadt, Germany, Peironcely, J. E.; Reijmers, T.; Coulier, L.; Bender, A.; Hankemeier, T. Understanding and classifying metabolite space and metabolite-likeness. PloS ONE 2011, 6, e Schrödinger Suite: Maestro, 9.2; Schrödinger: New York, NY, Pipeline Pilot Student Edition, version 6.1.5; Accelrys, Inc.: San Diego, CA, InChI, 1.03; International Union of Pure and Applied Chemistry (IUPAC): Research Triangle Park, NC, FIGURES Supporting Information Figure 1. Data processing scheme. Data preparation steps are indicated in orange, question 1 in black, question 2 in red, question 3 in green and question 4 in blue. Supporting Information Figure 2. MW/clogP distributions observed for (a) Approved Drugs, (b) HMDB and (c) TCM Database@Taiwan datasets. The non-clustered datasets are depicted in black and the clustered datasets in red/green/blue, respectively. (a) A strong preference of approved drugs for a combination of MW below 500 and a positive clogp < 5 is apparent. There is a trend for heavier approved drugs to be more lipophilic. (b) Human metabolites, regardless of their weight, populate the whole range of clogp values, from highly polar to highly apolar. (c) TCM molecules show a trend toward clogp values increasing with their MW. Supporting Information Figure 3. Number of (a) hydrogen bond acceptors, (b) hydrogen bond donors, (c) acidic and (d) basic atoms, (e) formal charge and (f) number of aromatic bonds per molecule for approved drugs (red), human metabolites (green) and TCM molecules (blue). For each color, the left bar represents the nonclustered and the right bar the clustered dataset. Supporting Information Figure 4. Molecular complexity of approved drugs (red), human metabolites (green) and TCM molecules (blue). For each color, the left bar represents the non-clustered and the right bar the clustered dataset. Approved drugs are of lower molecular complexity than human metabolites and TCM molecules. They have a lower average number of (a) rotatable bonds and (c) chiral centers. Also the average number of rings (b) is lower than for TCM molecules but higher than for human metabolites. Supporting Information Figure 5. Comparison of the MW and clogp distributions of molecules present in the non-clustered (continuous lines) and clustered (dashed lines) datasets. (a, d) Approved drugs, (b, e) human metabolites and (c, f) TCM molecules, with the respective datasets resulting from the overlap with the AMDB. The original datasets are plotted in red/green/blue and the overlapped datasets are depicted in black. 4

5 Supporting Information Figure 6. MW and clogp shifts observed for the clustered (a) approved drugs, (b) human metabolites and (c) TCM molecules. The top-level substrates are indicated in red/green/blue and their terminal metabolites are indicated in black. Terminal metabolites tend to be of higher MW and at the same time more hydrophilic. Supporting Information Figure 7. Shift in metabolite likeness introduced to (a) human metabolites and (b) TCM molecules by metabolism. Top-level substrates in green/blue, terminal metabolites in black for non-clustered (continuous lines) and clustered datasets (dashed lines). Supporting Information Figure 8. Comparison of shifts in MW and clogp of substrates and their metabolites induced by phase I (black) and phase II (red/green/blue) reactions, for (a, d) approved drugs, (b, e) human metabolites and (c, f) TCM molecules. The majority of phase I and phase II reactions increase the aqueous solubility of small organic molecules. Exceptions can be observed; these are often elimination reactions catalyzed by cytochrome P450s. Supporting Information Figure 9. Propensity of metabolic reactions resulting in metabolite with a >1 log unit increased clogp. Supporting Information Figure 10. MW/clogP distribution of metabolites in the bile (red), faeces (green) and urine (blue), for (a) approved drugs, (b) human metabolites and (c) TCM molecules. 5

6 Supporting Information Figure 1. 6

7 Molecular weight [Da] a 0 clogp clogp clogp TCM molecules 20 Human metabolites 20 Drugs Molecular weight [Da] b Molecular weight [Da] 1500 c Supporting Information Figure 2. 7

8 Fraction Drugs (red), h.metab. (green), TCM (blue) Fraction Drugs (red), h.metab. (green), TCM (blue) Fraction Drugs (red), h.metab. (green), TCM (blue) No. of hydrogen bond acceptors a No. of hydrogen bond donors b No. of acidic atoms c Fraction Drugs (red), h.metab. (green), TCM (blue) Fraction Drugs (red), h.metab. (green), TCM (blue) Fraction Drugs (red), h.metab. (green), TCM (blue) No. of basic atoms d Formal charge e No. of aromatic bonds f Supporting Information Figure 3. 8

9 Fraction Drugs (red), h.metab. (green), TCM (blue) Fraction Drugs (red), h.metab. (green), TCM (blue) Fraction Drugs (red), h.metab. (green), TCM (blue) No. of rotatable bonds a No. of rings b No. of chiral centres c Supporting Information Figure 4. 9

10 Drugs_original (red), drugs_overlapping (black) H.metab._original (green), h.metab._overlapping (black) TCM_original (blue), TCM_overlapping (black) Probability density Probability density Probability density Molecular weight [Da] a Molecular weight [Da] b Molecular weight [Da] c Drugs_original (red), drugs_overlapping (black) H.metab._original (green), h.metab._overlapping (black) TCM_original (blue), TCM_overlapping (black) Probability density Probability density Probability density clogp d clogp e clogp f Supporting Information Figure 5. 10

11 Drugs_parents (red), drugs_terminal metab. (black) H.metab._parents (green), h.metab._term. metab. (black) TCM_parents (blue), TCM_terminal metab. (black) clogp clogp clogp Molecular weight [Da] a Molecular weight [Da] b Molecular weight [Da] c Supporting Information Figure 6. 11

12 Parents (green), terminal metabolites (black) Parents (blue), terminal metabolites (black) Probability density Probability density Metabolite likeness a Metabolite likeness b Supporting Information Figure 7. 12

13 Metabolite molecular weight [Da] Drugs Phase I (black), drugs Phase II(red) Metabolite molecular weight [Da] h.metab. Phase I (black), h.metab. Phase II (green) Metabolite molecular weight [Da] TCM Phase I (black), TCM Phase II (blue) Substrate molecular weight [Da] a Substrate molecular weight [Da] b Substrate molecular weight [Da] c Drugs Phase I (black), drugs Phase II(red) H.metab. Phase I (black), h.metab. Phase II (green) TCM Phase I (black), TCM Phase II (blue) Metabolite clogp Metabolite clogp Metabolite clogp Substrate clogp d Substrate clogp e Substrate clogp f Supporting Information Figure 8. 13

14 0.5 Drugs (red), human metabolites (green), TCM (blue) 0.4 Fraction Dealkylation Oxidative_deamination Reduction(=O) OxidativeElimination Aromatization/Elimination Dephosphorylation Elimination Dehydration Reduction(=/ ) Oxidation( /=) Supporting Information Figure 9. 14

15 Bile (red), faeces (green), urine (blue) Bile (red), faeces (green), urine (blue) Bile (red), faeces (green), urine (blue) clogp clogp clogp Molecular weight [Da] a Molecular weight [Da] b Molecular weight [Da] c Supporting Information Figure

16 Supporting Information Table 1. Overview of Datasets Used for Statistical Analysis. Dataset Question 1: What physicochemical property space do approved drugs, human metabolites and molecules related to traditional Chinese medicine occupy? processed Approved Drugs of DrugBank 1391 processed & clustered Approved Drugs of DrugBank 876 processed HMDB 8355 processed & clustered HMDB 2792 processed TCM processed & clustered TCM processed AMDB: top-level substrates processed & clustered AMDB: top-level substrates 5431 Question 2: How does metabolism alter the physicochemical property space from substrate to metabolic product? processed AMDB top-level substrates comprised in the Approved Drugs of DrugBank 1341 processed AMDB top-level substrates comprised in the Approved Drugs of DrugBank, clustered 670 processed AMDB top-level substrates comprised in the HMDB 1144 processed AMDB top-level substrates comprised in the HMDB, clustered 408 processed AMDB top-level substrates comprised in the TCM 1131 processed AMDB top-level substrates comprised in the TCM clustered 429 Question 3: What shifts in physicochemical property space do individual metabolic reactions introduce into molecules? Phase I reactions recorded for AMDB metabolic schemes where the top-level substrate is comprised in the Approved Drugs of DrugBank, 2641 clustered Phase II reactions recorded for AMDB metabolic schemes where the top-level substrate is comprised in the Approved Drugs of DrugBank, 1228 clustered Phase I reactions recorded for AMDB metabolic schemes where the top-level substrate is comprised in the HMDB, clustered 1087 Phase II reactions recorded for AMDB metabolic schemes where the top-level substrate is comprised in the HMDB, clustered 714 Phase I reactions recorded for AMDB metabolic schemes where the top-level substrate is comprised in the TCM Database@Taiwan database, clustered Phase II reactions recorded for AMDB metabolic schemes where the top-level substrate is comprised in the TCM Database@Taiwan database, clustered Number of entries

17 Supporting Information Table 1 Continued. Question 4: What are the physicochemical properties of metabolites found in the bile, faeces and urine? metabolites in the bile recorded for AMDB metabolic schemes where the top-level substrate is comprised in the Approved Drugs of 782 DrugBank, clustered metabolites in the faeces recorded for AMDB metabolic schemes where the top-level substrate is comprised in the Approved Drugs of 948 DrugBank, clustered metabolites in the urine recorded for AMDB metabolic schemes where the top-level substrate is comprised in the Approved Drugs of 3063 DrugBank, clustered metabolites in the bile recorded for AMDB metabolic schemes where the top-level substrate is comprised in the HMDB, clustered 175 metabolites in the faeces recorded for AMDB metabolic schemes where the top-level substrate is comprised in the HMDB, clustered 102 metabolites in the urine recorded for AMDB metabolic schemes where the top-level substrate is comprised in the HMDB, clustered 970 metabolites in the bile recorded for AMDB metabolic schemes where the top-level substrate is comprised in the TCM Database@Taiwan database, clustered metabolites in the faeces recorded for AMDB metabolic schemes where the top-level substrate is comprised in the TCM Database@Taiwan database, clustered metabolites in the urine recorded for AMDB metabolic schemes where the top-level substrate is comprised in the TCM Database@Taiwan database, clustered

18 Supporting Information Table 2. Metabolic Reaction Types. Acetylation Acetylcysteination Acylation Alkylation Amide Hydrolysis Amination Aromatization/Elimination Azo cleavage Bromination Chlorination Condensation Cyanidation Cysteamination Dealkylation Dealkynylation Deamination Dehalogenation Dehydration Dehydrohalogenation Dehydroxylation Demethylation Denitration Dephosphorylation Desulfuration Elimination Epoxidation Epoxidation/Hydrolysis Epoxide dehydration Epoxide hydrolysis Epoxide opening Ester hydrolysis Esterification Formylation Glucosylation Glucuronidation Glutamation Glutathionylation Glycination Glycosylation Hydration Hydroxidation Hydroxylation Hydroxylation/Tautomerization Methiolation Methoxylation Methylation N-Dearylation N-dealkylation N2-elimination Nirosation Oxidation Oxidation(-/=) Oxidation/Dehalogenation Oxidative elimination Oxidative deamination Peroxidation Phosphorylation Rearrangement Reduction Reduction(=/-) Reduction(=O) Ring opening S-S Reduction Sulfation Sulfonation Sulfuration Tautomerization Thioester hydrolysis 18

19 Supporting Information Table 3. Physicochemical Properties and Descriptors of Molecules Included in the AMDB, Approved Drugs of DrugBank, HMDB and TCM property/descriptor d a mean d a d_c b sd d_c b sd h c mean h c h_c d sd h_c d sd t e mean t e t_c f sd t_c f sd mean mean mean molecular weight [Da] number of heavy atoms number of nitrogens number of oxygens number of halogens number of hydrogen bond acceptors number of hydrogen bond donors number of acidic atoms number of basic atoms formal charge number of hydrophobic atoms number of aromatic bonds number of rotatable bonds relative number of rotatable bonds number of rings number of chiral centers clogp logs vdw volume [Å 3 ] ASA [Å 2 ] pos. ASA [Å 2 ] neg. ASA [Å 2 ] ASA of hydrophobic atoms [Å 2 ] ASA of polar atoms (PSA) [Å 2 ] a approved drugs; b approved drugs, clustered dataset; c human metabolites; d human metabolites, clustered dataset; e TCM molecules; f TCM molecules, clustered dataset. 19

20 Supporting Information Table 4. Shifts in Physicochemical Property Space for Each Metabolic Scheme. property/descriptor d_shift a mean d_shift a sd d_c_shift b mean d_c_shift b sd h_shift c mean h_shift c sd h_c_shift d mean h_c_shift d sd t_shift e mean t_shift e sd t_c_shift f mean t_c_shift f sd molecular weight [Da] number of heavy atoms number of nitrogens number of oxygens number of hydrogen bond acceptors number of hydrogen bond donors number of acidic atoms number of basic atoms formal charge number of hydrophobic atoms number of aromatic bonds number of rotatable bonds relative number of rotatable bonds number of rings number of chiral centers clogp logs vdw volume [Å 3 ] ASA [Å 2 ] pos. ASA [Å 2 ] neg. ASA [Å 2 ] ASA of hydrophobic atoms [Å 2 ] ASA of polar atoms (PSA) [Å 2 ] a approved drugs; b approved drugs, clustered dataset; c human metabolites; d human metabolites, clustered dataset; e TCM molecules; f TCM molecules, clustered dataset. 20

21 Supporting Information Table 5. Shifts in Physicochemical Property Space for Each Phase I Reaction. property/descriptor d_c_shift a mean d_c_shift a sd h_c_shift b mean h_c_shift b sd t_c_shift c mean t_c_shift c sd molecular weight [Da] number of heavy atoms number of nitrogens number of oxygens number of halogens number of hydrogen bond acceptors number of hydrogen bond donors number of acidic atoms number of basic atoms formal charge number of hydrophobic atoms number of aromatic bonds number of rotatable bonds relative number of rotatable bonds number of rings number of chiral centers clogp logs vdw volume [Å 3 ] ASA [Å 2 ] pos. ASA [Å 2 ] neg. ASA [Å 2 ] ASA of hydrophobic atoms [Å 2 ] ASA of polar atoms (PSA) [Å 2 ] a approved drugs, clustered dataset; b human metabolites, clustered dataset; c TCM molecules, clustered dataset. 21

22 Supporting Information Table 6. Shifts in Physicochemical Property Space for Each Phase II Reaction. property/descriptor d_c_shift a mean d_c_shift a sd h_c_shift b mean h_c_shift b sd t_c_shift c mean t_c_shift c sd molecular weight [Da] number of heavy atoms number of nitrogens number of oxygens number of halogens number of hydrogen bond acceptors number of hydrogen bond donors number of acidic atoms number of basic atoms formal charge number of hydrophobic atoms number of aromatic bonds number of rotatable bonds relative number of rotatable bonds number of rings number of chiral centers clogp logs vdw volume [Å 3 ] ASA [Å 2 ] pos. ASA [Å 2 ] neg. ASA [Å 2 ] ASA of hydrophobic atoms [Å 2 ] ASA of polar atoms (PSA) a approved drugs, clustered dataset; b human metabolites, clustered dataset; c TCM molecules, clustered dataset. 22

23 Supporting Information Table 7a. Physicochemical Properties and Descriptors of Metabolites Found in the Bile. property/descriptor d_c_bile a mean d_c_bile a sd h_c_bile b mean h_c_bile b sd t_c_bile c mean t_c_bile c sd molecular weight [Da] number of heavy atoms number of nitrogens number of oxygens number of hydrogen bond acceptors number of hydrogen bond donors number of acidic atoms number of basic atoms formal charge number of hydrophobic atoms number of aromatic bonds number of rotatable bonds relative number of rotatable bonds number of rings number of chiral centers clogp logs vdw volume [Å 3 ] ASA [Å 2 ] pos. ASA [Å 2 ] neg. ASA [Å 2 ] ASA of hydrophobic atoms [Å 2 ] ASA of polar atoms (PSA) [Å 2 ] a approved drugs, clustered dataset; b human metabolites, clustered dataset; c TCM molecules, clustered dataset. 23

24 Supporting Information Table 7b. Physicochemical Properties and Descriptors of Metabolites Found in the Faeces. property/descriptor d_c_faeces a mean d_c_faeces a sd h_c_faeces b mean h_c_faeces b sd t_c_faeces c mean t_c_faeces c sd molecular weight [Da] number of heavy atoms number of nitrogens number of oxygens number of hydrogen bond acceptors number of hydrogen bond donors number of acidic atoms number of basic atoms formal charge number of hydrophobic atoms number of aromatic bonds number of rotatable bonds relative number of rotatable bonds number of rings number of chiral centers clogp logs vdw volume [Å 3 ] ASA [Å 2 ] pos. ASA [Å 2 ] neg. ASA [Å 2 ] ASA of hydrophobic atoms [Å 2 ] ASA of polar atoms (PSA) a approved drugs, clustered dataset; b human metabolites, clustered dataset; c TCM molecules, clustered dataset. 24

25 Supporting Information Table 7c. Physicochemical Properties and Descriptors of Metabolites Found in the Urine. property/descriptor d_c_urine a mean d_c_urine a sd h_c_urine b mean h_c_urine b sd t_c_urine c mean t_c_urine c sd molecular weight [Da] number of heavy atoms number of nitrogens number of oxygens number of hydrogen bond acceptors number of hydrogen bond donors number of acidic atoms number of basic atoms formal charge number of hydrophobic atoms number of aromatic bonds number of rotatable bonds relative number of rotatable bonds number of rings number of chiral centers clogp logs vdw volume [Å 3 ] ASA [Å 2 ] pos. ASA [Å 2 ] neg. ASA [Å 2 ] ASA of hydrophobic atoms [Å 2 ] ASA of polar atoms (PSA) [Å 2 ] a approved drugs, clustered dataset; b human metabolites, clustered dataset; c TCM molecules, clustered dataset. 25