General information about drugs
Drug Name:
Generic name of the drug
Drug class:
Classification of drugs based on therapeutic use and/or molecular targets
Brand name:
Brand name of the indicated drug.
TKI:
Tyrosine kinase inhibitor, includes small molecule TKIs and monoclonal antibodies against tyrosine kinases
Approval Date:
Approval date of the drug of interest as obtained from the FAERS database.
Human Drug Targets:
Human drug targets of the drug of interest as curated from DrugBank (Wishart et al 2018 and 2006 papers).
Cardiotoxic drug:
Drugs with a risk frequency of more than 1% in the category “Toxicity Cardiac” summarized by Herrman J 2020 were labeled as cardiotoxic. Drugs missing in this review were labeled cardiotoxic if there were associated with an Odds Ratio of at least 2 that we calculated using the FAERS database and the category “Cardiotoxicity”.
Adverse Events
Risk frequency:
This column shows the risk frequency curated from Herrman J 2020.
FAERS:
FDA Adverse Event Reporting System (https://www.fda.gov/drugs/drug-approvals-and-databases/fda-adverse-event-reporting-system-faers.).
Odds Ratio Rank (FAERS):
All TKIs, anthracyclines and proteasome inhibitors were ranked by decreasing Odds Ratio.
Odds Ratio (FAERS):
The odds ratio for a cardiotoxic adverse event that is associated with a drug of interest was calculated using data from the FAERS database.
Number of Years Analyzed:
This column indicates the number of years the FAERS data was collected that we used it for Odds Ratio calculation.
Number of All adverse events:
Total number of adverse events reported for a drug of interest in FAERS.
Number of cardiotoxic adverse events:
The number of cardiotoxic events for a drug of interest in FAERS.
Lower/Upper confidence interval:
Lower and upper confidence intervals for the calculated Odds Ratio for a drug of interest from data in FAERS
Cell line and Experimental Data/Metadata
IPSC-id:
Label of the iPSC-derived cardiomyocyte (CM) cell line.
Sex:
This column shows the sex of the donor of the iPSC-derived CM cell line.
Age (at biopsy):
This column shows the age of the donor at the time of skin biopsy.
Race/Ethnicity:
This column shows the race/ethnicity of the donor.
Dataset Characteristic:
Datasets were generated by bulk transcriptomics.
Cell Line:
The indicated cell line describes the iPSC-id that was stimulated with the indicated drug.
Bulk Transcriptomics:
Gene expression profiles of entire tissue samples or groups of cells that are extracted together to isolate mRNA for library preparation. Method used for this study.
Seq Platform:
The sequencing platform used for data generation.
Library Preparation:
All data was generated from randomly primed library samples.
Drug Stimulation Time:
Duration of the drug stimulation before cell harvesting in hours,i.e. 48 hours for all stimulations.
Drug Concentration:
Final concentration of the drug in the cell culture media.
Drug replicates count:
Numbers of biological replicates that were generated and subjected to bulk transcriptomic analysis for the indicated cell line/drug combination.
Control Replicates Count:
Number of biological replicates of each cell line that were treated with DMSO to generate control data.
DEGs:
Differential expressed genes represent genes whose levels are different between control and drug treated cell lines.
Drug Replicates Count Used for DEGs:
The quality control step of our RNAseq analysis pipeline removes outlier replicates before calculation of differentially expressed genes (Xiong Y et al. 2017). This number indicates how many biological replicates were finally subjected to differential expression analysis after quality control.
Control Replicates Count Used for DEGs:
Control replicates were subjected to the same quality control steps as drug replicates.
Drug induced DEGs in each cell line
NCBI gene symbol:
The human NCBI gene symbol of the up- or downregulated gene.
hg38:
Version of the human genome used in this study to identify genes, i.e. GRCH38.
Up or Down:
Specification of whether expression of a gene is up- or downregulated in the drug-treated cell line, compared to the control treated cell line.
Signed (+/-) -log10(p-value):
Minus log10(p-value) after removal of gene expression components that are not selective for the drug by our SVD-based pipeline (for details see Hansen J et al. 2023). To document the direction of change in these analyses we defined that up- and downregulated are associated with a positive and negative signed -log10(p-value), respectively.
SVD:
Singular Value Decomposition - an algebraic method to resolve complex data sets. In the gene expression data sets SVD allows for the resolution of the data set to identify genes associated with the response of interest.
Significance rank:
All up- and downregulated genes induced in the same cell line by the same drug were ranked by decreasing absolute signed -log10(p-value).
Drug-induced DEGs across all cell lines
Number of cell lines with upregulated gene:
Number of cell lines in which the gene of interest was upregulated and among the top 600 most significant genes after treatment with the indicated drug.
Number of cell lines with downregulated gene:
Number of cell lines in which the gene of interest was downregulated and among the top 600 most significant genes after treatment with the indicated drug.
Number of cell lines:
This column gives the number of cell lines treated with a drug of interest (after quality control by our pipeline (Xiong Y et al. 2017)).
% cell lines with gene among top 600 DEGs:
Percent of the treated cell lines a gene of interest was among the top 600 ranked genes calculated from the three preceding columns.
Drug-induced pathways in each cell line
All significant MBCO level-3 pathways (p-value ≤ 0.05) predicted for the top 600 averaged up- and down-regulated genes for each drug in each cell line.
Cardiomyocyte cell line:
The cell line of interest treated with the drug of interest.
For description of all other columns see above.
For description of all other columns see above.
Pathway Enrichment using Molecular Biology of the Cell Ontology (MBCO):
For pathway enrichment analysis we used the Molecular Biology of the Cell Ontology (MBCO, version 1.1), a focused cell biological ontology (Hansen at al. 2017, mbc-ontology.org). MBCO pathways typically span three to four levels, where the detail of the described biology increases with increasing level numbers. For the data presented here we used only level-3 pathways, since these pathways specify function at the gene level.
All significant MBCO level-3 pathways that were predicted for the top 600 up- and down-regulated genes after averaging the predicted DEGs for each drug across all cell lines are shown.
Up/Down:
This column documents if the pathway was predicted based on the up- or downregulated genes.
Pathway:
Name of the predicted MBCO level-3 pathway.
Enrichment p-value:
P-values were calculated by analyzing the significance of the overlap between all annotated pathway genes and up- or downregulated genes using Fisher’s exact test.
Enrichment rank:
For each drug pathways predicted from up- or downregulated genes were ranked by increasing p-values.
Pathway F1 score AUC:
If a cardiotoxic TKI up- or downregulates a pathway that was identified to be upregulated or downregulated, respectively, at higher ranks by cardiotoxic TKIs than by non-cardiotoxic TKIs, the calculated F1 score AUC for the pathway of interest is shown (for details see Hansen J et al. 2023).
Cardiotoxicity of pathway:
Based on the F1 AUC score, the pathway is up- or downregulated by cardiotoxic TKIs.
Pathway genes:
Within a pathway of interest all annotated genes that are up- or downregulated (as indicated in the column Up/Down) by a drug of interest are shown.
Drug Induced pathways associated with cardiotoxicity
Pathways:
Cardiotoxic pathway that is up- or downregulated at higher ranks by cardiotoxic TKIs as compared to non-cardiotoxic TKIs.
Up/Down:
A pathway is up- or downregulated by cardiotoxic TKIs.
Enrichment rank cutoff:
Pathway predicted from up- or downregulated genes induced by a drug of interest in a cell line of interest were ranked by increasing p-values. F1 score statistics were calculated at each enrichment rank cutoff, as described below.
Precision:
All cardiotoxic and non-cardiotoxic TKI treatments in any cell line that up- or downregulate a pathway of interest with an enrichment rank that is equal or lower to the enrichment rank cutoff are counted. Precision is the ratio of the counted cardiotoxic TKI treatments to total of counted cardiotoxic and non-cardiotoxic TKI treatments for pathway of interest.
Recall:
Indicates the ratio of the counted cardiotoxic TKI treatments for pathway of interest, as described above, to all cardiotoxic treatments.
F1 score (beta=0.25):
A weighted harmonic mean between the precision and recall gives the F1 score. Selecting the weighting factor beta as 0.25, the precision was emphasized over the recall, since we were interested in pathways that indicative of a cardiotoxic response.
F1 score AUC (corrected):
F1 scores over the ranks 1 to 30 were used to calculate the F1 score AUC rank. To emphasize identification of SCPs that are either up- or downregulated by cardiotoxic or non-cardiotoxic TKIs, we introduced a penalty if a pathway was identified as both up- and downregulated by cardiotoxic TKIs.
F1 score AUC rank:
Indicates the rank of all identified corrected F1 score AUCs for both up- and downregulated pathways.
Cardiotoxic TKIs:
List of all cardiotoxic TKIs that up- or downregulate a pathway of interest in the given cell line(s) at the given enrichment rank cutoff or a lower rank.
Non-cardiotoxic TKIs:
List of all non-cardiotoxic TKIs that up- or downregulate a pathway of interest in the given cell line(s) at the given enrichment rank cutoff or a lower rank.
Potential Genomic variants influencing drug PK or PD
Genomic variants that mapped to genes involved in a drug’s pharmacokinetic or pharmacodynamics, correlated with a transcriptomic outlier response in one cell line treated with that drug and had a population wide frequency of <=10% (for details see Hansen J et al. 2023). Drug target proteins, drug-metabolizing enzymes and drug transporters were obtained from DrugBank.
Variant RS ID:
Standard label for the variant in dbSNP or ClinVar.
Variant Gene:
Gene the genomic variant maps to.
Location of variant:
Genomic location of the genomic variant relative to the gene.
Relation of variant gene to drug target gene:
Characterization of the gene variant that indicates potential influence of the variant on proteins involved in the action of drug of interest.
Potentially affected drug-related protein:
Drug receptor, drug transporter or drug-metabolizing enzyme that is affected by the genomic variant. If the variant maps to any of these three proteins (i.e., Relation of variant gene to drug target genes is either ‘Drug target protein’ or ‘Enzyme’, ‘Transporter’), the entry is the same as in the column ‘Variant Gene’. However if the variant maps to a transcription factor (TF) or a protein kinase regulating any of these proteins, then the target of the regulatory protein is listed.
Potentially affected drug:
Drug that could be affected by the variant.
Chromosome, Start-position, End-position:
These three columns show the chromosomal position of the variant.
Potential allele:
Allele potentially influencing PK or PD of given drug.
Population-wide frequency:
Population-wide frequency of the potential allele.
Other allele(s):
This column shows the alternative alleles at the same position with a higher frequency.
Cardiomyocyte cell line with deviating response:
This cell line showed a transcriptomic response after treatment with the indicated drug that deviated from the transcriptomic response in all other treated cell lines and had a higher frequency of the selected variant than all other 5 cell lines.
Genotype of cardiomyocyte cell line:
This entry shows the genotype of the identified cell line described above.
Potential genomic variants influencing cardiotoxic pathways
Genomic variants mapping to pathways associated with TKI-triggered cardiotoxicity (for details see Hansen J et al. 2023).
Variant RS ID:
Standard label for the variant in dbSNP or ClinVar.
Variant Gene:
This entry shows the gene the variant maps to and that is annotated to the pathway of interest.
Pathway:
Cardiotoxic pathway that is up- or downregulated at higher ranks by cardiotoxic TKIs as compared to noncardiotoxic TKIs and contains the variant gene described above.
Up/Down:
A pathway is up- or downregulated by cardiotoxic TKIs.
Location of variant:
Genomic location of the genomic variant relative to the gene.
Chromosome, Start-position, End-position:
These three columns show the chromosomal position of the variant.
Potential allele:
Allele potentially influencing pathway activity.
Population-wide frequency:
Population-wide frequency of the potential allele.
Potentially relevant TKI(s):
The cardiotoxic TKIs that up- or downregulate the indicated pathway.
Arrhythmia MATLAB model
iPSC-CMs:
Cardiomyocytes (CM) derived from induced pluripotent stem cells (iPSC). Fibroblasts collected from healthy volunteers are reprogrammed into induced pluripotent stem cells, then the iPSCs are differentiated into ventricular cardiomyocytes that spontaneously beat. For details see Schaniel et al Stem Cell Reports (16(12):3036-3049. doi: 10.1016/j.stemcr.2021.10.005.).
Gene-model parameter relationship:
Mapping between drug-induced alterations in gene expression levels and changes in model parameters for simulations.
Action potential:
Waveform of cellular membrane voltage versus time. The increase in membrane voltage and subsequent decrease in voltage initiate the contraction and relaxation phases of the cardiac cycle. Action potentials in mature ventricular myocytes require electrical stimulation but occur spontaneously in most iPSC-CM preparations.
Calcium transients:
Transient elevation of intracellular calcium, and subsequent reduction, in response to action potentials.
Sarcomere length change:
Measure of iPSC-CM contraction that quantifies the distance between adjacent subcellular molecular landmarks. Decreases and increases in sarcomere length correspond to contraction and relaxation, respectively.
MSN02-4:
An IPSC-CM line obtained from subject 2 for details see Schaniel C et al. (2021).
MSN09-4:
An iPSC-CM line obtained from subject 9 for details see Schaniel C et al. (2021).
Hypertrophy MATLAB model
Hypertrophy model:
One main script and four dependent functions that together compute both the dynamic and the steady-state activity profiles of several intracellular signals associated with hypertrophy, or inappropriate cardiomyocyte growth.
Receptor inputs:
Proteins that respond directly to drugs, circulating hormones, or circulating signaling proteins. These are generally located in the cell membrane and can activate or inhibit additional signals within the cell.Classification of drugs based on therapeutic use and/or molecular targets
Brand name:
Brand name of the indicated drug.
Central pathways:
Different classical signaling pathways that control growth response in cardiomyocytes (for details see publications from Saucerman lab (Ryall et al. (2012) JBC 287. PMID: 23091058 Frank et al. (2018) JMCC: 121 PMID. 30030017).
Phenotypic output:
Changes in levels of six mRNAs associated with hypertrophy.
