4 Matrix and Tensor Factorization Methods …
71
angle, a large-scale analysis may even help us understand the different toxic states
of a cell and the molecular drivers of each cellular state.
While there are limitations in the current analysis, future extensions of the analyses
can advance our knowledge in various directions. First, using detailed drug-target
interactions in the models could help classify the on-target and off-target effects
more reliably; however, a key limitation here is to obtain large-scale standardized
drug-target profiles. Very recently works in standardizing the drug-target interactions
have come up on a large-scale [54] and exploring these for an integrated drug-targettoxicogenomic analysis would be an interesting future direction. Secondly, a large
majority of toxicity analysis is performed on data originating from cell line panels.
It would be valuable to explore if tissue-specific toxicity profiles are available for a
more robust and practically applicable analysis. Third, organism-level toxicity data
is limited to only a few organisms only; it is important to evaluate how comprehensive such modeling is in general and how widely the results can be applied across
organisms.
In terms of future developments in the toxicology practices, studies, and risk
assessment strategies, we hope the presented works could stimulate the integration
of advanced machine learning models. For example, the methods presented here can
be used to identify the markers of toxic response toward a data and knowledge-driven
approach for risk assessment.
References
1. Grabinger T et al (2014) Ex vivo culture of intestinal crypt organoids as a model system for
assessing cell death induction in intestinal epithelial cells and enteropathy. Cell Death Dis
5(5):e1228
2. Aberdam E et al (2017) Induced pluripotent stem cell-derived limbal epithelial cells (LiPSC)
as a cellular alternative for in vitro ocular toxicity testing. PLoS ONE 12(6):e0179913
3. Hartung T et al (2012) Food for thought … systems toxicology. ALTEX 29(2):119–128
4. Kohonen P et al (2017) A transcriptomics data-driven gene space accurately predicts liver
cytopathology and drug-induced liver injury. Nat Commun 8:15932
5. Kohonen P et al (2014) Cancer biology, toxicology and alternative methods development go
hand-in-hand. Basic Clin Pharmacol Toxicol 115:50–58
6. Grafström RC et al (2015) Toward the replacement of animal experiments through th
bioinformatics-driven analysis of ‘omics’ data from human cell cultures. Altern Lab Anim
43:325–332
7. Nymark P et al (2018) A data fusion pipeline for generating and enriching adverse outcome
pathway descriptions. Toxicol Sci 162(1):264–275
8. Yeakley JM et al (2017) A trichostatin a expression signature identified by TempO-Seq targeted
whole transcriptome profiling. PLoS One 12(5)
9. Costello JC et al (2014) A community effort to assess and improve drug sensitivity prediction
algorithms. Nat Biotechnol 32(12):1202–1212
10. Ammad-Ud-Din M et al (2014) Integrative and personalized QSAR analysis in cancer by
Kernelized Bayesian matrix factorization. J Chem Inf Model 54(8):2347–2359
11. Ammad-ud-din M et al (2016) Drug response prediction by inferring pathway-response associations with kernelized Bayesian matrix factorization. Bioinformatics 32(17):i455–i463
71
angle, a large-scale analysis may even help us understand the different toxic states
of a cell and the molecular drivers of each cellular state.
While there are limitations in the current analysis, future extensions of the analyses
can advance our knowledge in various directions. First, using detailed drug-target
interactions in the models could help classify the on-target and off-target effects
more reliably; however, a key limitation here is to obtain large-scale standardized
drug-target profiles. Very recently works in standardizing the drug-target interactions
have come up on a large-scale [54] and exploring these for an integrated drug-targettoxicogenomic analysis would be an interesting future direction. Secondly, a large
majority of toxicity analysis is performed on data originating from cell line panels.
It would be valuable to explore if tissue-specific toxicity profiles are available for a
more robust and practically applicable analysis. Third, organism-level toxicity data
is limited to only a few organisms only; it is important to evaluate how comprehensive such modeling is in general and how widely the results can be applied across
organisms.
In terms of future developments in the toxicology practices, studies, and risk
assessment strategies, we hope the presented works could stimulate the integration
of advanced machine learning models. For example, the methods presented here can
be used to identify the markers of toxic response toward a data and knowledge-driven
approach for risk assessment.
References
1. Grabinger T et al (2014) Ex vivo culture of intestinal crypt organoids as a model system for
assessing cell death induction in intestinal epithelial cells and enteropathy. Cell Death Dis
5(5):e1228
2. Aberdam E et al (2017) Induced pluripotent stem cell-derived limbal epithelial cells (LiPSC)
as a cellular alternative for in vitro ocular toxicity testing. PLoS ONE 12(6):e0179913
3. Hartung T et al (2012) Food for thought … systems toxicology. ALTEX 29(2):119–128
4. Kohonen P et al (2017) A transcriptomics data-driven gene space accurately predicts liver
cytopathology and drug-induced liver injury. Nat Commun 8:15932
5. Kohonen P et al (2014) Cancer biology, toxicology and alternative methods development go
hand-in-hand. Basic Clin Pharmacol Toxicol 115:50–58
6. Grafström RC et al (2015) Toward the replacement of animal experiments through th
bioinformatics-driven analysis of ‘omics’ data from human cell cultures. Altern Lab Anim
43:325–332
7. Nymark P et al (2018) A data fusion pipeline for generating and enriching adverse outcome
pathway descriptions. Toxicol Sci 162(1):264–275
8. Yeakley JM et al (2017) A trichostatin a expression signature identified by TempO-Seq targeted
whole transcriptome profiling. PLoS One 12(5)
9. Costello JC et al (2014) A community effort to assess and improve drug sensitivity prediction
algorithms. Nat Biotechnol 32(12):1202–1212
10. Ammad-Ud-Din M et al (2014) Integrative and personalized QSAR analysis in cancer by
Kernelized Bayesian matrix factorization. J Chem Inf Model 54(8):2347–2359
11. Ammad-ud-din M et al (2016) Drug response prediction by inferring pathway-response associations with kernelized Bayesian matrix factorization. Bioinformatics 32(17):i455–i463