Artificial neural network analysis of data from multiple in vitro assays for prediction of skin sensitization potency of chemicals

Hirota, Morihiko; Kouzuki, Hirokazu; Ashikaga, Takao; Sono, Sakiko; Tsujita, Kyoko; Sasa, Hitoshi; Aiba, Setsuya

doi:10.1016/j.tiv.2013.02.013

Cited by 30 publications

(50 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Applications of machine learning methods to the construction of DAs are Bayesian networks (BN) (Jaworska et al, 2011;Jaworska et al, 2013), Artificial Neural Networks (ANN) (Hirota et al, 2013;Tsujita-Inoue et al, 2014;Hirota et al, 2015;Tsujita-Inoue et al, 2015), Naïve…”

Section: Machine Learning Approachesmentioning

confidence: 99%

“…"2 out of 3" ITS approach: (Bauch et al, 2012;Natsch et al, 2013;Urbisch et al, 2015a); Kao ITS and Kao STS: (Nukada et al, 2013;Takenouchi et al, 2015); RIVM STS:(van der Veen et al, 2014a;van der Veen et al, 2014b): Stacking meta model: (Gomes, 2012); IDS: (Matheson, 2015;Strickland et al, 2016); BN ITS: Jaworska et al, 2013;Jaworska et al, 2015); ANN ITS: (Hirota et al, 2013;Tsujita-Inoue et al, 2014;Hirota et al, 2015); EC-JRC: (Dimitrov et al, 2005;Asturiol et al, 2016); Global and local regression models: ; IATA: (Patlewicz et al, 2014, Patlewicz et al, 2015 was checked using the leave-one-out validation (Strickland et al, 2016). For the ANN-ITS, the ability of the model to predict the final decision outcome on the skin sensitisation potential was validated using the 10-fold cross validation approach (Hirota et al, 2013). Probabilistic approaches for the assessment of potency or potential offer the statistical tools to combine information of different parameters such as the BN ITS with the use of M...…”

Section: Balancing Information Gains and Costsmentioning

confidence: 99%

“…the RIVM STS, reported direct testing cost estimates (van der Veen et al 2014). In case of probabilistic DAs Matheson et al 2015;Hirota et al 2015;Hirota et al 2013;TsujitaInoue et al 2015), it is indicated that due to the saving unnecessary testing will also save costs. Furthermore, the Bayesian network (Jaworska et al 2013) and the "2 out of 3" ITS (Bauch et al 2012;Urbisch et al 2015a) suggest that a non-animal testing method would be avoided if information collected as a certain step of the strategy would be sufficient to conclude on the skin sensitisation potential.…”

Section: Balancing Information Gains and Costsmentioning

confidence: 99%

“…The overall conclusion is also based on a majority vote from test results from sequential steps in the strategy. The Artificial Neural Network (ANN) concept Hirota et al 2013;Tsujita-Inoue et al 2015) or the Bayesian networks Jaworska and Hoffmann, 2010;Jaworska et al, 2013;Jaworska et al, 2015) offer probabilistic approaches to predict skin sensitisation potential or potency using different physicochemical properties and information from nonanimal testing methods. The IDS DA (Matheson, 2015;Strickland et al, 2016) uses different machine learning approaches, i.e.…”

Section: Balancing Information Gains and Costsmentioning

confidence: 99%

“…Hence, they are presented in different ways, for example in the form of qualitative flowcharts (Grindon et al, 2008e;Mekenyan et al, 2010;ECHA, 2014a;Patlewicz et al, 2015), probabilistic approaches (machine learning) applying Artificial Neural Networks (Hirota et al, 2013;Tsujita-Inoue et al, 2014;Hirota et al, 2015) or Bayesian Networks (Jaworska and Hoffmann, 2010;Jaworska et al, 2013;Jaworska et al, 2015), and as deterministic approaches based on a "majority vote" decision rule for batteries of testing methods (Bauch et al, 2012;van der Veen et al, 2014a;van der Veen et al, 2014b;Urbisch et al, 2015a) or scorebased batteries of testing methods (Ellison et al, 2010;Nukada et al, 2013;Takenouchi et al, 2015). In addition, a regression analysis model ) and a quantitative model using the toxico-kinetics and toxico-dynamics modelling (MacKay et al, 2013) are used.…”

mentioning

confidence: 99%

See 4 more Smart Citations

An economic approach to non-animal toxicity testing for skin sensitisation

Leontaridou¹

View full text Add to dashboard Cite

Section: Machine Learning Approachesmentioning

confidence: 99%

Section: Balancing Information Gains and Costsmentioning

confidence: 99%

Section: Balancing Information Gains and Costsmentioning

confidence: 99%

Section: Balancing Information Gains and Costsmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

An economic approach to non-animal toxicity testing for skin sensitisation

Leontaridou¹

View full text Add to dashboard Cite

Systemic drugs inducing non‐immediate cutaneous adverse reactions and contact sensitizers evoke similar responses in THP‐1 cells

Gonçalo

Martins

Silva

et al. 2014

J of Applied Toxicology

View full text Add to dashboard Cite

Contact sensitizers induce phenotypic and functional changes in dendritic cells (DC) that enhance their antigen-presenting capacity and, ultimately, modulate the T cell response. To evaluate if there is a similar effect of drugs causing T-cell-mediated cutaneous adverse drug reactions (CADR), we studied the in vitro effect of drugs on THP-1 cells, a cell line widely used to evaluate the early molecular and cellular events triggered by contact sensitizers. The effect of allopurinol, oxypurinol, ampicillin, amoxicillin, carbamazepine and sodium valproate, at EC30 concentrations, was evaluated on p38 MAPK activation, by Western Blot, and on the expression of genes coding for DC maturation markers, pro-inflammatory cytokine/chemokines and hemeoxygenase 1 (HMOX1), by real-time RT-PCR. Results were compared with lipopolysaccharide (LPS), a DC maturation stimulus, and the strong contact sensitizer, 1-fluoro-2,4-dinitrobenzene (DNFB). All drugs studied significantly upregulated HMOX1 gene transcription and all, except the anticonvulsants, also upregulated IL8. Allopurinol and oxypurinol showed the most intense effect, in a magnitude similar to DNFB and superior to betalactams. Transcription of CD40, IL12B and CXCL10 genes by drugs was more irregular. Moreover, like DNFB, all drugs activated p38 MAPK, although significantly only for oxypurinol. Like contact sensitizers, drugs that cause non-immediate CADR activate THP-1 cells in vitro, using different signalling pathways and affecting gene transcription with an intensity that may reflect the frequency and severity of the CADR they cause. Direct activation of antigen-presenting DC by systemic drugs may be an important early step in the pathophysiology of non-immediate CADR.

show abstract

Evaluation of combinations of in vitro sensitization test descriptors for the artificial neural network‐based risk assessment model of skin sensitization

Hirota

Fukui²,

Okamoto³

et al. 2015

J of Applied Toxicology

Self Cite

View full text Add to dashboard Cite

The skin sensitization potential of chemicals has been determined with the use of the murine local lymph node assay (LLNA). However, in recent years public concern about animal welfare has led to a requirement for non-animal risk assessment systems for the prediction of skin sensitization potential, to replace LLNA. Selection of an appropriate in vitro test or in silico model descriptors is critical to obtain good predictive performance. Here, we investigated the utility of artificial neural network (ANN) prediction models using various combinations of descriptors from several in vitro sensitization tests. The dataset, collected from published data and from experiments carried out in collaboration with the Japan Cosmetic Industry Association (JCIA), consisted of values from the human cell line activation test (h-CLAT), direct peptide reactivity assay (DPRA), SH test and antioxidant response element (ARE) assay for chemicals whose LLNA thresholds have been reported. After confirming the relationship between individual in vitro test descriptors and the LLNA threshold (e.g. EC3 value), we used the subsets of chemicals for which the requisite test values were available to evaluate the predictive performance of ANN models using combinations of h-CLAT/DPRA (N = 139 chemicals), the DPRA/ARE assay (N = 69), the SH test/ARE assay (N = 73), the h-CLAT/DPRA/ARE assay (N = 69) and the h-CLAT/SH test/ARE assay (N = 73). The h-CLAT/DPRA, h-CLAT/DPRA/ARE assay and h-CLAT/SH test/ARE assay combinations showed a better predictive performance than the DPRA/ARE assay and the SH test/ARE assay. Our data indicates that the descriptors evaluated in this study were all useful for predicting human skin sensitization potential, although combinations containing h-CLAT (reflecting dendritic cell-activating ability) were most effective for ANN-based prediction.

show abstract

Artificial neural network analysis of data from multiple in vitro assays for prediction of skin sensitization potency of chemicals

Cited by 30 publications

References 36 publications

An economic approach to non-animal toxicity testing for skin sensitisation

An economic approach to non-animal toxicity testing for skin sensitisation

Systemic drugs inducing non‐immediate cutaneous adverse reactions and contact sensitizers evoke similar responses in THP‐1 cells

Evaluation of combinations of in vitro sensitization test descriptors for the artificial neural network‐based risk assessment model of skin sensitization

Contact Info

Product

Resources

About