Supplementary Material
A high-throughput metabolomics method to predict high concentration cytotoxicity of drugs from low concentration profiles
Stéphanie Heux1,3, Thomas J. Fuchs2,3, Joachim Buhmann2,3, Nicola Zamboni1,3, Uwe Sauer1,3
1 Institute of Molecular Systems Biology, ETH Zurich, Wolfgang-Pauli Strasse 16, 8093 Zurich, Switzerland
2 Department of Computer Science, ETH Zurich, Universitätstrasse 6, 8092 Zurich, Switzerland
3Competence Center for Systems Physiology and Metabolic Diseases, Schafmattstrasse 18, 8093 Zurich, Switzerland
Corresponding Author: Uwe Sauer, ,Phone +41 44 633 36 72
Supplementary Fig. 1: Growth response to drug dosage
Maximal specific growth rate of cells in h-1 plotted against increasing drug dosage. The concentrations in mg/L are listed in Supplementary Table 2. Plotted is the average from quadruplicate experiments. In total, seven concentrations were used for each drug. The horizontal red line shows the concentration at which the curve kinked, i.e. where the growth rate dropped.
1
Supplementary Fig. 2: 13C metabolic flux analysis of drug-treated cells.
Distribution of seven independently determined flux ratios in 82 (41 compounds at 2 concentrations) conditions. In each case, the median of the distribution is indicated by a vertical line, the 25th percentile by the grey box and the 90th percentile by the horizontal line. Data points outline of the 90th percentile are indicated by dots and name of the corresponding condition is given. Names flagged with (*) correspond to the highest drug concentration. The untreated controls are indicated by open circles.Abbreviations: 2DG, 2-deoxyglucose, 5FU, 5-flurouracil.
Supplementary Fig. 3: Reproducibility of high-throughput metabolomics platform
Reproducibility both within- and between-day precision assays and biological variability, both intra- and inter-plate for yeast extracts. Only data for the 55 metabolites with acceptable linearity and reproducibility are provided. Dash lines indicate the average relative standard deviation for each condition tested i.e. intra-day, repeated injections of one cell extract the same day; inter-day, repeated injections of one cell extract over two days; intra-plate, injection of 3 different cells extract obtained from the same plate; inter-plate, injection of 3 different cells extract obtained from three different plates. Metabolites numbers are defined in supplementary Table 5.
Supplementary Fig. 4: Metabolome differences of yeast cells challenged with presumably metabolically inert compounds.
The set included nonspecific chemicals (e.g. vitamins, antibacterial agents, neurotransmitters and buffer) and various on-metabolized carbon source (i.e. xylose, arabinose, sorbitol, raffinose). Columns represent the two concentrations (L=0.5 g/l and H=5g/l) used for the different compounds. Data represent the average of 3 independent cultures per condition. Fold change is relative to the cells grown on glucose only. The statistical significance of the observed relative metabolite changes was assessed using t-test. Only metabolites showing a significant change at the p < 0.01 levels are shown.
Supplementary Fig. 5: Intracellular metabolites level at the xylose node assimilation.
Depicted is the mean area (n= 3 independent samples) ± SD of metabolites contents between 10 g/l glucose grown yeast cells (1) ; 0.5 g/l xylose/10 g/l glucose grown yeast cells (2) and 5 g/l xylose/10 g/l glucose grown yeast cells (3).Values significantly different from the glucose grown yeast cells as determined by a t-test are marked with one asterisk (p < 0.05) or two asterix (p < 0.01). The graph was created by using visualisation system Vanted (Junker et al. 2006). As expected, most of the tested compounds remain inert to the cells. Surprisingly xylose, on which S. cerevisiae cannot grow, induces significant changes (p<0.01) (Supplementary Fig. 4). These changes were localised around the endogenous xylose assimilation pathway and a linear relationship was observed between the magnitude of the changes and the amount of xylose added in the medium (Supplementary Fig. 4), indicating a co-metabolism of the xylose together with the glucose.
Supplementary Fig. 6: Cross validation experiments for assessing prediction accuracy of the kNN classifier.
Class balanced misclassification error for toxicity prediction per frame. Learning a kNN classifier based on all frames from all drugs yields a training error of 9%. The generalization error is estimated with 24% in a cross validation experiment, for which in turn all frames of a drug were left out, then the model was learned on the remaining frames and subsequently the left out frames were predicted by the model. A permutation test was conducted to test the whole learning and feature selection procedure for overfitting. To this end the toxicity labels of the frames were randomly permuted and the complete learning procedure was conducted to train a model on the random labels. Repeating this experiment 100 times yields an average classification error of 51% demonstrating that the proposed classification model does not overfit.
Supplementary Table 1: List of compounds tested
The set comprised 41 small bioactive molecules drug with 74 drug target interactions. The drug target interactions were extracted from the Drug bank database(Wishart et al. 2008; Wishart et al. 2006), the STITCH database(Kuhn et al. 2008), the BRENDA database(Chang et al. 2009; Schomburg et al. 2002). In order to study putative target proteins that are associated with the main mechanism of action, we report only the interactions that are at least mentioned two times in the 3 different databases or that have experimental evidence of direct chemical-protein binding (i.e. binding affinity (Ki) derived from the BRENDA database). In addition, we excluded metabolizing enzymes that do not have homologous in yeast (e.g. Cytochrome P450 2C9), multidrug resistance-associated proteins (e.g. ABC Proteins family) and unspecific binders (e.g. albumin, retroviral envelope proteins). The set is composed of 10 compounds with reported interactions with metabolic process (central carbohydrates metabolism), 27 therapeutic agents used against metabolic diseases with a wide range of targets (lipid metabolism, signalling network, ions transporters and DNA replication), 4 antifungal drugs which target the lipid and amino acids metabolism and the microtubule and one stimulant affecting a variety of cellular process.
1
CAS Number / Chemical Compound / FDA Approved / Target name / Target protein name (Human) / Target protein name (Yeast) / Target function / Pharmacological Action / Therapeutic area75330-75-5 / Lovastatin / yes / 3-Hydroxy-3-methylglutaryl-coenzyme A reductase / HMGCR / HMG1, HMG2 / Lipid Metabolism / Anticholesteremic Agents / Hypercholesterolemia
81131-70-6 / Pravastatin / yes / 3-Hydroxy-3-methylglutaryl-coenzyme A reductase / HMGCR / HMG1, HMG2 / Lipid Metabolism
Solute carrier organic anion transporter family member 1B1 / SLCO1B1 / none / Ions Transporter
25812-30-0 / Gemfibrozil / yes / Peroxisome proliferator activated receptor alpha / PPARA / none / Signalling Network / Antilipemic Agents
Lipoprotein lipase precursor / LPL / TGL2 / Lipid Metabolism
49562-28-9 / Fenofibrate / yes / Peroxisome proliferator activated receptor alpha / PPARA / none / Signalling Network
97322-87-7 / Troglitazone / Withdrawn / Peroxisome proliferator activated receptor gamma / PPARG / none / Signalling Network / Hypoglycemic Agents / Diabetes
Plasminogen activator inhibitor 1 / SERPINE1 / none / Signalling Network
Long-chain-fatty-acid--CoA ligase 4 / ACSL4 / FAA1, FAA2, FAA3, FAA4, FAT1, FAT2 / Lipid Metabolism
74772-77-3 / Ciglitizone / Withdrawn / Peroxisome proliferator activated receptor gamma / PPARG / none / Signalling Network
1115-70-4 / Metformin / yes / 5'-AMP-activated protein kinase subunit beta-1 / PRKAB1 / SNF1 / Signalling Network
834-28-6 / Phenformin / Withdrawn / 5'-AMP-activated protein kinase subunit beta-1 / PRKAB1 / SNF1 / Signalling Network
94-20-2 / Chlorpropamide / yes / ATP-sensitive inward rectifier potassium channel 1 / KCNJ1 / none / Ions Transporter
64-77-7 / Tolbutamide / yes / ATP-sensitive inward rectifier potassium channel 2 / KCNJ2 / none / Ions Transporter
6-phosphofructo-2-kinase/fructose-2,6-biphosphatase / PFKB1, PFKB2, PFKB3, PKKB4 / FBP26 / Alternative carbon source metabolism
135062-02-1 / Repaglinide / yes / ATP-sensitive inward rectifier potassium channel 1 / KCNJ1 / none / Ions Transporter
96829-58-2 / Orlistat / yes / triacylglycerol lipase precursor / PNLIP / TGL2 / Lipid Metabolism / Anti-Obesity Agents / Obesity
Lipoprotein lipase precursor / LPL / TGL2 / Lipid Metabolism
28395-03-1 / Bumetanide / yes / Solute carrier family 12 member 4 / SLC12A4 / YBR235W / Ions Transporter / Diuretics / Hypertension
Solute carrier family 12 member 1 / SLC12A1 / YBR235W / Ions Transporter
Solute carrier family 12 member 2 / SLC12A2 / YBR235W / Ions Transporter
54-31-9 / Furosemide / yes / Solute carrier family 12 member 1 / SLC12A1 / YBR235W / Ions Transporter
Sodium/potassium-transporting ATPase alpha-1 chain precursor / ATP1A1 / PMR1, ENA1, ENA2, ENA5, PMC1, PMA1, PMA2 / Ions Transporter
58-94-6 / Chlorothiazide / yes / Carbonic anhydrase 1 / CA1 / NCE103 / Nitrogen metabolism
Carbonic anhydrase 2 / CA2 / NCE103 / Nitrogen metabolism
Carbonic anhydrase 4 precursor / CA4 / NCE103 / Nitrogen metabolism
Solute carrier family 12 member 3 / SLC12A3 / YBR235W / Ions Transporter
Calcium-activated potassium channel alpha subunit 1 / KCNMA1 / none / Ions Transporter
26807-65-8 / Indapamide / yes / Potassium voltage-gated channel subfamily E member 1 / KCNE1 / none / Ions Transporter
Potassium voltage-gated channel subfamily KQT member 1 / KCNQ1 / none / Ions Transporter
21829-25-4 / Nifedipine / yes / Dihydropyridine-sensitive L-type, calcium channel alpha-2/delta subunits precursor / CACNA2D1 / none / Ions Transporter / Antihypertensive Agents
66085-59-4 / Nimodipine / yes / Voltage-dependent calcium channel gamma-1 subunit / CACNG1 / none / Ions Transporter
15663-27-1 / Cisplatin / yes / Deoxyribonucleic acid / DNA / DNA / DNA replication / Antineoplastic Agents / Cancer
50-76-0 / Actinomycin D / yes / Deoxyribonucleic acid / DNA / DNA / DNA replication / Antibiotics, Antineoplastic
23214-92-8 / Doxorubicin / yes / DNA topoisomerase 2-alpha / TOP2A / TOP2 / DNA replication
147-94-4 / Ara-CMP / yes / DNA polymerase alpha catalytic subunit / POLA / POL1, CDC2 / DNA replication / Immunosuppressive Agents
51-21-8 / 5-Fluorouacil / yes / Dihydropyrimidine dehydrogenase [NADP+] precursor / DPYD / none / DNA replication
Thymidylate synthase / TYMS / CDC21 / Nucleotide metabolism
59-05-2 / Methotrexate / yes / Dihydrofolate reductase / DHFR / DFR1 / Nucleotide metabolism / Antimetabolites, Antineoplastic
Folylpolyglutamate synthase, mitochondrial precursor / FPGS / RMA1 / Nucleotide metabolism
Thymidylate synthase / TYMS / CDC21 / Nucleotide metabolism
127-07-1 / Hydroxyurea / yes / Ribonucleoside-diphosphate reductase large subunit / RRM1 / RNR2, RNR4 / Nucleotide metabolism / Antineoplastic Agents
Ribonucleoside-diphosphate reductase M2 subunit / RRM2 / RNR2, RNR4 / Nucleotide metabolism
25387-67-1 / Camptothecin / experimental / DNA topoisomerase I / TOP1 / TOP2 / DNA replication
10540-29-1 / Tamoxifen / yes / Estrogen receptor / ESR1 / none / Signalling Network
Cholestenol DELTA-isomerase / EBP / ERG2 / Lipid Metabolism
154-17-6 / 2-Deoxyglucose / experimental / Hexokinase / HK / HXK1, HXK2 / Central carbon Metabolism / nd / Candidate for the treatment of cancer (Warburg effect)
129-46-4 / Suramin / yes / phosphoglycerate kinase / PGK / PGK1 / Central carbon Metabolism / Antinematodal Agents / Treatment of African trypanosomiasis
DNA topoisomerase 2-beta / TOP2B / TOP2 / Nucleotide metabolism
288-13-1 / Pyrazole / experimental / Alcohol dehydrogenase E chain / ADHFE1 / ADH1 / Central carbon Metabolism / Enzyme Inhibitors / none
97-77-8 / Disulfiram / yes / Aldehyde dehydrogenase, mitochondrial precursor / ALDH2 / ALD2-6 / Central carbon Metabolism / Alcohol Deterrents / Treatment of alcoholism
Aldehyde dehydrogenase X, mitochondrial precursor / ALDH1B1 / ALD2-6 / Central carbon Metabolism
Aldehyde dehydrogenase, dimeric NADP-preferring / ALDH3A1 / ALD2-6 / Central carbon Metabolism
Retinal dehydrogenase 1 / ALDH1A1 / none / Central carbon Metabolism
Dopamine beta-hydroxylase precursor / DBH / none / Signalling Network
Peripheral-type benzodiazepine receptor / BZRP / none / Signalling Network
305-53-3 / Iodoacetate / not listed / Glyceraldehyde-3-phosphate dehydrogenase, liver / GAPDH / TDH1, TDH2, TDH3 / Central carbon Metabolism / Alkylating Agents / Candidate for the treatment of cancer (Warburg effect)
Glyceraldehyde-3-phosphate dehydrogenase, testis-specific / HSD35 / TDH1, TDH2, TDH3 / Central carbon Metabolism
614-05-1 / Oxythiamine / not listed / Pyruvate dehydrogenase / PDHA1 / PDA1, PDB1 / Central carbon Metabolism / Antimetabolites / Candidate for the treatment of cancer (Warburg effect)
Transketolase / TK / TKL1, TKL2 / Central carbon Metabolism
Thiamin pyrophosphokinase 1 / TPK1 / THI80 / Vitamins and Cofactors Metabolism
141-82-2 / Malonate / experimental / Succinate dehydrogenase [ubiquinone] iron-sulfur protein, mitochondrial precursor / SDHB / SDH1, SDH2, SDH3, SDH4 / Central carbon Metabolism
329-89-5 / 6-Aminonicotinamide / not listed / 6-Phosphogluconate dehydrogenase, decarboxylating / PGD / GND1, GND2 / Central carbon Metabolism / Teratogens / Candidate for the treatment of cancer (Warburg effect)
Oligomycin / not listed / H+-transporting two-sector ATPase / ATP Family / ATP6, ATP8, OLI1 / Energy Metabolism / Anti-Bacterial Agents / none
74222-97-2 / Sulfometuron methyl / not listed / ilvB (bacterial acetolactate synthase)-like isoform 1 / ILVBL / ILV2 / Amino acids metabolism / Herbicides / none
23593-75-1 / Clotrimazole / yes / Sterol 14-demethylase / CYP51A1 / ERG11 / Lipid Metabolism / Antifungal Agents / Treatment of fungal infections
Intermediate conductance calcium-activated potassium channel protein 4 / KCNN4 / KCNN4 / Ions Transporter
6-phosphofructokinase / PFK-L, PFK-M / PFK1, PFK2 / Central carbon Metabolism
67564-91-4 / Fenpropimorph / not listed / Cholestenol DELTA-isomerase / EBP / ERG2 / Lipid Metabolism / Industrial fungicide / none
178-04-35-2 / Benomyl / not listed / Beta-tubulin and alpha-tubulin / TBA, TBB, TTLL / TUB1, TUB2, TUB3, TUB4 / Cell-division cycle / Industrial fungicide / none
Mitotic spindle assembly checkpoint / MAD2L1 / MAD2 / Cell-division cycle
Glucose-6-phosphate isomerase / GPI / PGI1 / Central carbon Metabolism
58-08-2 / Caffeine / yes / Adenosine A1 receptor / ADORA1 / none / Signalling Network / Central Nervous System Stimulants / Treatement of migraine , headache ,drowsiness and mild water-weight gain
Adenosine A2a receptor / ADORA2A / none / Signalling Network
3',5'-cyclic-nucleotide phosphodiesterase / PDE4B / PDE1 / Signalling Network
1
Supplementary Table 2: Drug treatment concentrations
Dose index / 0 / 1 / 2 / 3 / 4 / 5 / 6 / 7Chemical Compound / Concentration (mg/L)
Lovastatin / 0 / 0.015 / 0.15 / 1.5 / 7.5 / 15 / 75 / 150
Pravastatin / 0 / 0.02 / 0.2 / 2 / 10 / 20 / 100 / 200
Gemfibrozil / 0 / 0.00075 / 0.0075 / 0.075 / 0.375 / 0.75 / 3.75 / 7.5
Fenofibrate / 0 / 0.015 / 0.15 / 1.5 / 7.5 / 15 / 75 / 150
Troglitazone / 0 / 0.007 / 0.07 / 0.7 / 3.5 / 7 / 35 / 75
Ciglitizone / 0 / 0.04 / 0.4 / 4 / 20 / 40 / 200 / 400
Metformin / 0 / 0.1 / 1 / 10 / 50 / 100 / 500 / 1000
Phenformin / 0 / 0.2 / 2 / 20 / 100 / 200 / 1000 / 2000
Chlorpropamide / 0 / 0.1 / 1 / 10 / 50 / 100 / 500 / 1000
Tolbutamide / 0 / 0.015 / 0.15 / 1.5 / 7.5 / 15 / 75 / 150
Repaglinide / 0 / 0.05 / 0.5 / 5 / 25 / 50 / 250 / 500
Orlistat / 0 / 0.05 / 0.5 / 5 / 25 / 50 / 250 / 500
Bumetanide / 0 / 0.05 / 0.5 / 5 / 25 / 50 / 250 / 500
Furosemide / 0 / 0.05 / 0.5 / 5 / 25 / 50 / 250 / 500
Chlorothiazide / 0 / 0.075 / 0.75 / 7.5 / 37.5 / 75 / 375 / 750
Indapamide / 0 / 0.075 / 0.75 / 7.5 / 37.5 / 75 / 375 / 750
Nifedipine / 0 / 0.015 / 0.15 / 1.5 / 7.5 / 15 / 75 / 150
Nimodipine / 0 / 0.016 / 0.16 / 1.6 / 8 / 16 / 80 / 160
Cisplatin / 0 / 0.003 / 0.03 / 0.3 / 1.5 / 3 / 15 / 30
Actinomycin D / 0 / 0.002 / 0.002 / 0.2 / 1 / 2 / 10 / 20
Doxorubicin / 0 / 0.0006 / 0.006 / 0.06 / 0.3 / 0.6 / 3 / 6
Ara-CMP / 0 / 0.125 / 1.25 / 12.5 / 62.5 / 125 / 625 / 1250
5FU / 0 / 0.000006 / 0.00006 / 0.0006 / 0.003 / 0.006 / 0.03 / 0.06
Methotrexate / 0 / 0.00005 / 0.0005 / 0.005 / 0.025 / 0.05 / 0.25 / 0.5
Hydroxyurea / 0 / 0.3 / 3 / 30 / 150 / 300 / 1500 / 3000
Camptothecin / 0 / 0.02 / 0.2 / 2 / 10 / 20 / 100 / 200
Tamoxifen / 0 / 0.002 / 0.002 / 0.2 / 1 / 2 / 10 / 20
2-Deoxyglucose / 0 / 0.015 / 0.15 / 1.5 / 7.5 / 15 / 75 / 150
Suramin / 0 / 0.1 / 1 / 10 / 50 / 100 / 500 / 1000
Pyrazole / 0 / 0.025 / 0.25 / 2.5 / 12.5 / 25 / 125 / 250
Disulfiram / 0 / 0.00005 / 0.0005 / 0.005 / 0.025 / 0.05 / 0.25 / 0.5
Sodium Iodoacetate / 0 / 0.0003 / 0.003 / 0.03 / 0.15 / 0.3 / 1.5 / 3
Oxythiamine / 0 / 0.15 / 1.5 / 15 / 75 / 150 / 750 / 1500
Malonate / 0 / 4 / 40 / 400 / 2000 / 4000 / 20000 / 40000
6-Aminonicotinamide / 0 / 0.14 / 1.4 / 14 / 70 / 140 / 700 / 1400
Oligomycin / 0 / 0.005 / 0.05 / 0.5 / 2.5 / 5 / 25 / 50
Sulfometuron methyl / 0 / 0.00001 / 0.0001 / 0.001 / 0.005 / 0.01 / 0.05 / 0.1
Clotrimazole / 0 / 0.00012 / 0.0012 / 0.012 / 0.06 / 0.12 / 0.6 / 1.2
Fenpropimorph / 0 / 0.00015 / 0.0015 / 0.015 / 0.075 / 0.15 / 0.75 / 1.5
Benomyl / 0 / 0.000125 / 0.00125 / 0.0125 / 0.0625 / 0.125 / 0.625 / 1.25
Caffeine / 0 / 0.1 / 1 / 10 / 50 / 100 / 500 / 1000
Supplementary Table 3: Global recovery of different metabolites
Intracellular metabolites content of samples obtained using the filter plate and shake flask cultivation (see Methods) were quantify using a GC-TOF according to the method as previously described (Ewald et al. 2009). The recovery was calculated based on the concentration of each metabolite according to the equation: recovery = (CFP/CSF)* 100%, where CFP is the concentration of the metabolite measured in the filter plate sample, and CSF is the concentration of the metabolite measured in the shake flask. When the recovery results were higher than 100%, the recovery was presented as > 100%. Recoveries above 100 % were typically observed when high metabolite concentrations were extracted from the biomass. In general, the filter plate approach ensure a high recovery for all the tested metabolites compare with a standard procedure, except for the glucose-6-phosphate.This can be explained by a longer quenching step when using the filter plate method; 50s maximum instead of 15s for the standard protocol.
Metabolites / Recovery (%)Glucose-6-phosphate / 38.3
Fructose-6-phosphate / 75.5
Fructose-1,6-bisphosphate / 99.5
Glycerol-phosphate / 90.6
Phosphoenol-pyruvate / 100.0
Citrate / 85.2
Fumarate / 99.5
Malate / 99.2
6-phosphogluconate / 76.2
Ribose-5-phosphate / 72.8
Alanine / 75.9
Aspartate / >100
Glutamate / >100
Glycine / >100
Histidine / 100
Lysine / >100
Methionine / 88.5
Phenylalanine / 100
Proline / 77.6
Serine / >100
Threonine / >100
Tyrosine / 100
Valine / >100
Supplementary Table 4: List of the metabolites detected
Amino acids / Carbohydrate derivatives/precursorsAlanine / 1,3-Bis-phosphoglycerate*
Arginine / 3-Phosphoglycerate/2-Phosphoglycerate
Aspartate / 6-Phospho-d-gluconate
Glutamate / d-Hexose-P
Glutamine / Dihydroxy-acetone-P,/Glyceraldehyde-3-P*
Histidine / d-Rib(ul)ose-5-P/Xylose-5-P
Lysine / Erythrose-4-P
Methionine / Fructose-1,6-bis-P
Phenylalanine / Glucosamine-P
Proline / Glyceraldehyde*
Serine / Glycerol-3-P
Tryptophan / Phosphoenolpyruvate
Tyrosine / Trehalose/Disaccahrides
Valine / Glucuronate*
Amino acid derivatives/precursors / Sedoheptulose-7-P
Citrulline / Pentose
Ornithine / Hexoses
Phenylpyruvate / Vitamins and derivatives
Nucleoside bases / Nicotinate
Adenine / Pantothenate
Guanine / Carboxylic acids
Nucleoside monophosphates / Citrate
AMP / Malate
Cyclic-AMP* / Succinate*
GMP / Alpha-Ketoglutarate*
Cyclic-GMP* / Pyruvate
Nucleoside di/triphosphates / Glyoxylate*
ADP / Redox-electron-carriers and precursors
ATP/ dGTP* / FAD*
CTP / FMN
dCTP / NAD+*
GTP / NADH
GDP / NADPH
Nucleotide precursors/derivatives / Oxidized glutathione
N-Acetyl-glucosamine* / Oxidized glutathione
CoA's / Reduced glutathione
Acetyl-CoA / Lipid precursors/derivatives
CoA / myo-Inositol
Malonyl-CoA / Mevalonate
Succinyl-CoA
These 68 compounds correspond to 15 % of the metabolome of yeast evaluated from the genome-wide metabolic reconstruction (Blank et al. 2005). They were selected from the library as previously developed (Buscher et al. 2009). The list of compounds includes 14 of the proteinogenic amino acids, 12 nucleotides, 23 components of the central carbohydrate metabolism and almost all redox cofactors.Namesflagged with (*) correspond to metabolites excluded due to a poor signal to noise and yielded to erratic data (see supplementary Table 5).Abbreviations: P, phosphate. A collective name or two names are given for isomers that cannot be differentiated by product ion.
Supplementary Table 5: Method linearity for yeast extract
Metabolite / Metabolite number / Signal to noise ratio / R2 / Splope3-Phosphoglycerate, 2-Phosphoglycerate / 1 / 14.9 / 0.91 / 6.60E+04
6-Phospho-d-gluconate / 2 / 53.5 / 0.99 / 6.96E+04
Acetyl-CoA / 3 / 33.7 / 1.00 / 1.89E+04
Adenine / 4 / 24.8 / 0.93 / 6.55E+04
ADP / 5 / 50.9 / 0.97 / 3.41E+04
Alanine / 6 / 18.0 / 0.98 / 1.81E+05
AMP / 7 / 128.0 / 1.00 / 1.82E+05
Arginine / 8 / 93.6 / 0.99 / 2.71E+05
Aspartate / 9 / 447.6 / 0.99 / 7.64E+05
Citrate / 10 / 77.1 / 0.98 / 4.00E+06
Citrulline / 11 / 1215.9 / 1.00 / 5.96E+06
CoA / 12 / 31.6 / 1.00 / 2.43E+04
CTP / 13 / 20.7 / 0.98 / 1.48E+04
dCTP / 14 / 34.7 / 0.96 / 6.85E+03
Erythrose-4-P / 15 / 18.7 / 0.98 / 4.72E+04
Fructose-1,6-bis-P / 16 / 34.9 / 0.99 / 1.70E+05
FMN / 17 / 23.1 / 0.95 / 1.84E+04
GDP / 18 / 62.6 / 0.92 / 1.72E+04
Glucosamine-6-P/Glucoseamine-1-P / 19 / 79.2 / 0.98 / 1.24E+04
Glutamate / 20 / 138.4 / 0.98 / 6.98E+06
Glutamine / 21 / 29.3 / 1.00 / 2.12E+06
Glycerol-3-P / 22 / 42.6 / 0.96 / 1.63E+05
GMP / 23 / 167.8 / 1.00 / 4.75E+04
GTP / 24 / 43.5 / 0.96 / 6.26E+03
Oxidised glutathione / 25 / 56.7 / 0.98 / 3.49E+04
Reduced glutathione / 26 / 77.9 / 0.97 / 1.98E+05
Guanine / 27 / 521.7 / 0.95 / 5.70E+03
Hexoses / 28 / 398.9 / 0.98 / 5.91E+06
d-Hexose-P / 29 / 54.2 / 0.99 / 1.44E+05
Histidine / 30 / 101.2 / 0.94 / 2.84E+05
Lysine / 31 / 25.7 / 1.00 / 1.22E+06
Malate / 32 / 35.5 / 0.98 / 9.67E+05
Malonyl-CoA / 33 / 25.9 / 0.94 / 1.75E+04
Methionine / 34 / 426.7 / 0.99 / 3.98E+03
Mevalonate / 35 / 25.2 / 0.94 / 1.02E+05
myo-Inositol / 36 / 29.4 / 1.00 / 4.51E+05
NADH / 37 / 58.4 / 0.98 / 2.66E+04
NADPH / 38 / 74.2 / 0.95 / 2.28E+04
Nicotinate / 39 / 135.6 / 0.97 / 1.64E+05
Ornithine / 40 / 26.7 / 0.99 / 4.00E+06
Pantothenate / 41 / 140.9 / 0.99 / 2.44E+05
Pentose / 42 / 118.0 / 0.96 / 7.97E+04
Phosphoenolpyruvate / 43 / 18.8 / 0.93 / 7.01E+04
Phenylalanine / 44 / 35.0 / 0.99 / 5.14E+04
Phenylpyruvate / 45 / 51.4 / 0.96 / 9.51E+04
Proline / 46 / 24.8 / 0.93 / 7.13E+03
Pyruvate / 47 / 24.0 / 0.91 / 2.94E+04
d-Rib(ul)ose-5-P / 48 / 31.6 / 0.98 / 9.34E+04
Sedoheptulose-7-P / 49 / 82.9 / 0.98 / 1.04E+05
Serine / 50 / 36.1 / 0.96 / 1.44E+05
Succinyl-CoA / 51 / 35.6 / 0.91 / 6.87E+03
Trehalose/Disaccahrides / 52 / 14.9 / 0.99 / 9.40E+04
Tryptophan / 53 / 19.1 / 0.95 / 3.95E+03
Tyrosine / 54 / 36.7 / 0.97 / 1.38E+04
Valine / 55 / countless / 0.99 / 5.56E+04
Metabolites excluded
1,3-Bis-phosphoglycerate / 0.5 / 0.73 / 1.45E+04
á-Ketoglutarate / 1.2 / 0.75 / -1.32E+06
ATP/dGTP / 4.2 / 0.12 / -5.29E+02
Cyclic-AMP / 0.6 / 0.73 / 1.01E+04
Cycic-GMP / 1.4 / 0.27 / -3.76E+02
FAD / 1.2 / 0.01 / 1.36E+02
Dihydroxy-acetone-P/Glyceraldehyde-3-P / 2.6 / 0.82 / 1.31E+05
N-Acetyl-glucosamine / 2.8 / 0.78 / 1.95E+03
Glucuronate / 1.0 / 0.45 / 2.07E+04
Glyceraldehyde / 2.5 / 0.94 / 2.77E+05
Glyoxylate / 0.7 / 1.00 / 9.61E+04
NAD+ / 2.1 / 0.99 / 4.55E+04
Succinate / 2.0 / 0.93 / 1.05E+05
The signal to noise ratio is defined by C/N where C is the compound peak area of an extract sample and N is compound peak area of a water sample. Out of 68 detected metabolites, 13 metabolites are associated with a poor signal to noise and yielded to erratic data.
Supplementary Table 6: Classification error rates with specificity and sensitivity
Class balanced misclassification error rate per frame for training, leave one drug out cross validation and for the permutation test. While the classifier is more sensitive to cytotoxicity during training, the cross validation experiments show a well-balanced behaviour.
Training / Leave One Drug OutCross Validation / Permutation Test
Balanced Error Rate / 0.0879 / 0.2401 / 0.5036 ± 0.0575
Sensitivity / 0.9492 / 0.7966 / 0.6701 ± 0.0832
Specificity / 0.8750 / 0.7232 / 0.3226 ± 0.0599
References
Blank, L. M., Kuepfer, L., & Sauer, U. (2005). Large-scale 13C-flux analysis reveals mechanistic principles of metabolic network robustness to null mutations in yeast. Genome Biology, 6(6), R49.
Buscher, J. M., Czernik, D., Ewald, J. C., Sauer, U., & Zamboni, N. (2009). Cross-Platform Comparison of Methods for Quantitative Metabolomics of Primary Metabolism. Analytical chemistry, 81(6), 2135-2143.
Chang, A., Scheer, M., Grote, A., Schomburg, I., & Schomburg, D. (2009). BRENDA, AMENDA and FRENDA the enzyme information system: new content and tools in 2009. Nucleic acids research, 37(Database issue), D588-592.
Ewald, J. C., Heux, S., & Zamboni, N. (2009). High-throughput quantitative metabolomics: workflow for cultivation, quenching, and analysis of yeast in a multiwell format. Analytical chemistry, 81(9), 3623-3629.
Junker, B. H., Klukas, C., & Schreiber, F. (2006). VANTED: a system for advanced data analysis and visualization in the context of biological networks. BMC Bioinformatics, 7, 109.
Kuhn, M., von Mering, C., Campillos, M., Jensen, L. J., & Bork, P. (2008). STITCH: interaction networks of chemicals and proteins. Nucleic acids research, 36(Database issue), D684-688.
Schomburg, I., Chang, A., & Schomburg, D. (2002). BRENDA, enzyme data and metabolic information. Nucleic acids research, 30(1), 47-49.
Wishart, D. S., Knox, C., Guo, A. C., Cheng, D., Shrivastava, S., Tzur, D., et al. (2008). DrugBank: a knowledgebase for drugs, drug actions and drug targets. Nucleic acids research, 36(Database issue), D901-906.
Wishart, D. S., Knox, C., Guo, A. C., Shrivastava, S., Hassanali, M., Stothard, P., et al. (2006). DrugBank: a comprehensive resource for in silico drug discovery and exploration. Nucleic acids research, 34(Database issue), D668-672.
1