Archives of Toxicology

, Volume 93, Issue 4, pp 931–940 | Cite as

Transfer learning for predicting human skin sensitizers

  • Chun-Wei TungEmail author
  • Yi-Hui Lin
  • Shan-Shan Wang


Computational prioritization of chemicals for potential skin sensitization risks plays essential roles in the risk assessment of environmental chemicals and drug development. Given the huge number of chemicals for testing, computational methods enable the fast identification of high-risk chemicals for experimental validation and design of safer alternatives. However, the development of robust prediction model requires a large dataset of tested chemicals that is usually not available for most toxicological endpoints, especially for human data. A small training dataset makes the development of effective models difficult with insufficient coverage and accuracy. In this study, an ensemble tree-based multitask learning method was developed incorporating three relevant tasks in the well-defined adverse outcome pathway (AOP) of skin sensitization to transfer shared knowledge to the major task of human sensitizers. The results show both largely improved coverage and accuracy compared with three state-of-the-art methods. A user-friendly prediction server was available at As AOPs for various toxicity endpoints are being actively developed, the proposed method can be applied to develop prediction models for other endpoints.


Adverse outcome pathway Allergic contact dermatitis Alternative method Multitask learning Skin sensitization ExtraTrees 



This work was supported by Ministry of Science and Technology of Taiwan [MOST104-2221-E-037-001-MY3, MOST107-2221-E-037-005-MY3]; National Health Research Institutes [NHRI-107A1-EMCO-0318184]; and Research Center for Environmental Medicine in Kaohsiung Medical University from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan. This work was initiated in Kaohsiung Medical University and completed in Taipei Medical University. The funding agencies play no role in the study design, data analysis and manuscript preparation.

Compliance with ethical standards

Conflict of interest

The authors declare no conflict of interest.

Supplementary material

204_2019_2420_MOESM1_ESM.docx (26 kb)
Supplementary material 1 (DOCX 25 KB)


  1. Alves V, Muratov E, Capuzzi S et al (2016a) Alarms about structural alerts. Green Chem 18(16):4348–4360. CrossRefGoogle Scholar
  2. Alves VM, Capuzzi SJ, Muratov E et al (2016b) QSAR models of human data can enrich or replace LLNA testing for human skin sensitization. Green Chem 18(24):6501–6515. CrossRefGoogle Scholar
  3. Basketter DA, Alepee N, Ashikaga T et al (2014) Categorization of chemicals according to their relative human skin sensitizing potency. Dermatitis 25(1):11–21. CrossRefGoogle Scholar
  4. Basketter DA, White IR, McFadden JP, Kimber I (2015) Skin sensitization: Implications for integration of clinical data into hazard identification and risk assessment. Hum Exp Toxicol 34(12):1222–1230. CrossRefGoogle Scholar
  5. Bauch C, Kolle SN, Ramirez T et al (2012) Putting the parts together: combining in vitro methods to test for skin sensitizing potentials. Regul Toxicol Pharmacol 63(3):489–504. CrossRefGoogle Scholar
  6. Bossa C, Benigni R, Tcheremenskaia O, Battistelli CL (2018) (Q)SAR methods for predicting genotoxicity and carcinogenicity: scientific rationale and regulatory frameworks. Methods Mol Biol 1800:447–473. CrossRefGoogle Scholar
  7. Braga RC, Alves VM, Muratov EN et al (2017) Pred-Skin: A Fast and Reliable Web Application to Assess Skin Sensitization Effect of Chemicals. J Chem Inf Model 57(5):1013–1017. CrossRefGoogle Scholar
  8. Dionisio KL, Frame AM, Goldsmith MR et al (2015) Exploring consumer exposure pathways and patterns of use for chemicals in the environment. Toxicol Rep 2:228–237. CrossRefGoogle Scholar
  9. Ezendam J, Braakhuis HM, Vandebriel RJ (2016) State of the art in non-animal approaches for skin sensitization testing: from individual test methods towards testing strategies. Arch Toxicol 90(12):2861–2883. CrossRefGoogle Scholar
  10. Geurts P, Ernst D, Wehenkel L (2006) Extremely randomized trees. Mach Learn 63(1):3–42. CrossRefGoogle Scholar
  11. Golla S, Madihally S, Robinson RL Jr, Gasem KA (2009) Quantitative structure-property relationship modeling of skin sensitization: a quantitative prediction. Toxicol In Vitro 23(3):454–465. CrossRefGoogle Scholar
  12. Guha R (2007) Chemical informatics functionality in R. J Stat Softw 18(5):1–16CrossRefGoogle Scholar
  13. Hanser T, Barber C, Marchaland JF, Werner S (2016) Applicability domain: towards a more formal definition. SAR QSAR Environ Res 27(11):893–909. CrossRefGoogle Scholar
  14. Huang SH, Tung CW (2017) Identification of consensus biomarkers for predicting non-genotoxic hepatocarcinogens. Sci Rep 7:41176. CrossRefGoogle Scholar
  15. Jaworska J, Hoffmann S (2010) Integrated Testing Strategy (ITS) - Opportunities to better use existing data and guide future testing in toxicology. ALTEX 27(4):231–242CrossRefGoogle Scholar
  16. Jaworska JS, Natsch A, Ryan C, Strickland J, Ashikaga T, Miyazawa M (2015) Bayesian integrated testing strategy (ITS) for skin sensitization potency assessment: a decision support system for quantitative weight of evidence and adaptive testing strategy. Arch Toxicol 89(12):2355–2383. CrossRefGoogle Scholar
  17. Jin H, Ling CX (2005) Using AUC and accuracy in evaluating learning algorithms. IEEE Trans Knowl Data Eng 17(3):299–310. CrossRefGoogle Scholar
  18. Karlberg AT, Bergstrom MA, Borje A, Luthman K, Nilsson JL (2008) Allergic contact dermatitis–formation, structural requirements, and reactivity of skin sensitizers. Chem Res Toxicol 21(1):53–69. CrossRefGoogle Scholar
  19. Kleinstreuer NC, Hoffmann S, Alepee N et al (2018) Non-animal methods to predict skin sensitization (II): an assessment of defined approaches (*). Crit Rev Toxicol 48(5):359–374. CrossRefGoogle Scholar
  20. Lepoittevin JP (2006) Metabolism versus chemical transformation or pro- versus prehaptens? Contact Dermatitis 54(2):73–74. CrossRefGoogle Scholar
  21. Li Y, Pan D, Liu J et al (2007a) Categorical QSAR Models for skin sensitization based upon local lymph node assay classification measures part 2: 4D-fingerprint three-state and two-2-state logistic regression models. Toxicol Sci 99(2):532–544. CrossRefGoogle Scholar
  22. Li Y, Tseng YJ, Pan D et al (2007b) 4D-fingerprint categorical QSAR models for skin sensitization based on the classification of local lymph node assay measures. Chem Res Toxicol 20(1):114–128. CrossRefGoogle Scholar
  23. Miller MD, Yourtee DM, Glaros AG, Chappelow CC, Eick JD, Holder AJ (2005) Quantum mechanical structure-activity relationship analyses for skin sensitization. J Chem Inf Model 45(4):924–929. CrossRefGoogle Scholar
  24. Netzeva TI, Worth A, Aldenberg T et al (2005) Current status of methods for defining the applicability domain of (quantitative) structure-activity relationships. The report and recommendations of ECVAM Workshop 52. Altern Lab Anim 33(2):155–173Google Scholar
  25. OECD (2018) OECD QSAR Toolbox, Accessed 13 Sep 2018
  26. Patlewicz GY, Basketter DA, Pease CK et al (2004) Further evaluation of quantitative structure–activity relationship models for the prediction of the skin sensitization potency of selected fragrance allergens. Contact Dermatitis 50(2):91–97. CrossRefGoogle Scholar
  27. Patlewicz G, Jeliazkova N, Safford RJ, Worth AP, Aleksiev B (2008) An evaluation of the implementation of the Cramer classification scheme in the Toxtree software. SAR QSAR Environ Res 19(5–6):495–524. CrossRefGoogle Scholar
  28. Patlewicz G, Casati S, Basketter DA et al (2016) Can currently available non-animal methods detect pre and pro-haptens relevant for skin sensitization? Regul Toxicol Pharmacol 82:147–155. CrossRefGoogle Scholar
  29. Roberts DW, Api AM (2018) Chemical applicability domain of the local lymph node assay (LLNA) for skin sensitisation potency. Part 4. Quantitative correlation of LLNA potency with human potency. Regul Toxicol Pharmacol 96:76–84. CrossRefGoogle Scholar
  30. Roberts DW, Patlewicz G (2018) Non-animal assessment of skin sensitization hazard: Is an integrated testing strategy needed, and if so what should be integrated? J Appl Toxicol 38(1):41–50. CrossRefGoogle Scholar
  31. Rogers D, Hahn M (2010) Extended-connectivity fingerprints. J Chem Inf Model 50(5):742–754. CrossRefGoogle Scholar
  32. Sarath Kumar KL, Tangadpalliwar SR, Desai A, Singh VK, Jere A (2016) Integrated Computational Solution for Predicting Skin Sensitization Potential of Molecules. PLoS One 11(6):e0155419. CrossRefGoogle Scholar
  33. Simm J, Magrans De Abril I, Sugiyama M (2014) Tree-Based Ensemble Multi-Task Learning Method for Classification and Regression. IEICE Transactions on Information Systems E97D(6):1677–1681
  34. Sushko I, Salmina E, Potemkin VA, Poda G, Tetko IV (2012) ToxAlerts: a Web server of structural alerts for toxic chemicals and compounds with potential adverse reactions. J Chem Inf Model 52(8):2310–2316. CrossRefGoogle Scholar
  35. Tung CW (2013) Prediction of pupylation sites using the composition of k-spaced amino acid pairs. J Theor Biol 336:11–17. CrossRefGoogle Scholar
  36. Tung CW, Ho SY (2008) Computational identification of ubiquitylation sites from protein sequences. BMC Bioinformatics 9:310. CrossRefGoogle Scholar
  37. Tung CW, Wang CC, Wang SS (2018) Mechanism-informed read-across assessment of skin sensitizers based on SkinSensDB. Regul Toxicol Pharmacol 94:276–282. CrossRefGoogle Scholar
  38. Urbisch D, Mehling A, Guth K et al (2015) Assessing skin sensitization hazard in mice and men using non-animal test methods. Regul Toxicol Pharmacol 71(2):337–351. CrossRefGoogle Scholar
  39. Wang CC, Lin YC, Wang SS, Shih C, Lin YH, Tung CW (2017) SkinSensDB: a curated database for skin sensitization assays. J Cheminform 9:5. CrossRefGoogle Scholar
  40. Willighagen EL, Mayfield JW, Alvarsson J et al (2017) The Chemistry Development Kit (CDK) v2.0: atom typing, depiction, molecular formulas, and substructure searching. J Cheminform 9(1):33. CrossRefGoogle Scholar
  41. Yuan H, Huang J, Cao C (2009) Prediction of skin sensitization with a particle swarm optimized support vector machine. Int J Mol Sci 10(7):3237–3254. CrossRefGoogle Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Authors and Affiliations

  1. 1.Graduate Institute of Data Science, College of ManagementTaipei Medical UniversityTaipeiTaiwan
  2. 2.National Institute of Environmental Health SciencesNational Health Research InstitutesMiaoli CountyTaiwan
  3. 3.School of PharmacyKaohsiung Medical UniversityKaohsiungTaiwan

Personalised recommendations