Application of Bayesian networks for hazard ranking of nanomaterials to support human health risk assessment

Figures & data

Figure 1. The method steps.

Table 1. Grouping of endpoints for the determination of the different biological effects.

Biological effects	Possible endpoints for the determination of the biological effect
Neurological effects	Neuropathology; motor activity; motor and sensory function; learning and memory; neurotransmitters; (OECD, 2007)
Reaches central nervous system	“Is found in the brain” as element (Morgan, 2005)
Cardiopulmonary effects	Lung or heart inflammation (inflammatory cell infiltration); lung hyper responsiveness; heart rate (variability); blood pressure; alterations in red blood cells; Broncho alveolar lavage fluid analysis (cytokines, LDH and protein) (Chuang et al., 2014; Gordon et al., 1998)
Immunological effects	Immunosuppression; increased propensity for allergic disease, including atopy, food allergies and asthma; hypersensitivity reactions directed at the chemical itself; Increased risk of autoimmune disease; dysfunctional responses of innate immune cells producing tissue or organ damage or dysfunction (WHO, 2012)
Inflammation	Systemic inflammation (white blood cells, lymphocytes and granulocytes); tissue inflammation; inflammatory cell infiltration; cytokines level (Chuang et al., 2014)
Fibrosis	Histopathological fibrosis analysis Liver: periportal fibrosis (incomplete septa); bridging fibrosis with lobular distortion; cirrhosis (Hiramatsu et al., 2006) Lung: thickening of alveolar or bronchiolar vessels; damage to lung structure; formation of fibrous bands or fibrous masses (Ashcroft et al., 1988)
Cytotoxicity	Cytotoxicity; necrosis; cell viability; apoptosis
Genotoxicity	Single-strand and double-strand breaks, mutations, deletions, chromosomal aberrations, impairment in DNA repair and cell-cycle (Ng et al., 2010)

Table 2. Criteria for the classification of the biological effects.

Classification effect		Criteria			Effect levelScore (BEik)
None	Endpoint(s) that fall under the defined effects were tested	No significant difference in tested endpoint compared to control			0
Low	Endpoint(s) that fall under the defined effects were tested	Significant difference in testedendpoint compared to control	Positive for one/few of the tests indicating this effect or in few animals		1
Medium	Endpoint(s) that fall under the defined effects were tested	Significant difference in testedendpoint compared to control	Dose-responserelationship	Positive for several of the tests indicating this effect or in several animals	2
				<75% decrease in cell viability
High	Endpoint(s) that fall under the defined effects were tested	Significant difference in testedendpoint compared to control	Dose-responserelationship	Positive for most of the tests indicating this effect or in most animals	3
				>75% decrease in cell viability

Figure 2. A typical Bayesian network showing the probabilistic relationship between a set of nodes.

Figure 3. NM characteristics mentioned by national experts.

Figure 4. NM characteristics mentioned by international experts.

Figure 5. Distribution of answers to the question related to the states of node “Dissolution distribution”.

Figure 6. The BN structure containing state probabilities of each node, where the green ellipses are physicochemical properties of NM, orange ellipses are biological effects, yellow ellipses are additional nodes for administration routes and study type and finally ellipse is the NM hazard node. CNS: (Reaches) central nervous system.

Table 3. Validation results of the BN; input values (20 cases).

Input parameters												NM hazard
N	Studytype	Shape	Nanoparticle	Dissolution	Surfacearea	Surfacecharge	Surfacecoatings	Surfacereactivity	Aggregation	Particlesize	Administrationroute	Observed	Predicted
1	In vivo	Sphere	Ag	–	–	–	–	–	–	10– 50	Inhalation	High	High
2	In vitro	Amorph	Al2O3	0–25%	–	–	–	–	–	10–50	–	None	None
3	In vivo	Sphere	Au	–	0–15	−25–0	SO3–	–	–	10–50	Inhalation	None	None
4	In vitro	Elongated	C	–	189–2025	–	–	–	–	>100	–	Medium	Low
5	In vitro	Elongated	Cellulose	–	–	–	Sulfate	–	Low	>100	–	Low	Low
6	In vivo	Irregular	CeO2	–	15–51	–	–	–	Medium	10–50	Inhalation	High	High
7	In vitro	Amorph	Co3O4	0–25%	–	–	–	High	–	10–50	–	High	High
8	In vitro	Sphere	CoO	0–25%	–	–	–	High	–	50–100	–	High	High
9	In vitro	Sphere	Cr2O3	0–25%	–	–	–	Low	–	>100	–	Medium	Medium
10	In vitro	Sphere	CuO	0–25%	–	–	–	High	–	10–50	–	High	High
11	In vivo	Sphere	Fe2O3		–	0–25	–	–	Medium	50–100	Oral	None	None
12	In vitro	Sphere	Fe3O4	0–25%	–	–	–	–	–	10–50	–	None	None
13	In vitro	Sphere	Ni2O3	0–25%	–	–	–	High	–	>100	–	Medium	Medium
14	In vitro	Sphere	NiO	0–25%	–	–	–	–	–	10–50	–	None	None
15	In vivo	Sphere	PS		–	25–50	Amine	–	–	50–100	Oral	None	Low
16	In vitro	Amorph	Sb2O3	0–25%	–	–		–	–	10–50	–	None	None
17	In vitro	Irregular	SiO2	–	189–2025	−50 to −25	Carbon	Low	High	10–50	–	Low	High
18	In vivo	Irregular	TiO2	–	15–51	–	–	–	High	>100	Oral	Medium	Medium
19	In vitro	Sphere	WO3	0–25%	–	–	–	–		10–50	–	None	None
20	In vitro	Elongated	ZnO	–	0–15	0 to 25	None	–	Medium	>100	–	High	High

Table 4. BN results; biological effects prediction.

	Cytotoxicity		Pulmonary effects		Fibrosis		RCNS effects		Genotoxicity		Inflammation
N	Observed	Predicted	Observed	Predicted	Observed	Predicted	Observed	Predicted	Observed	Predicted	Observed	Predicted
1	–	–	High	High	–	–	–	–	–	–	Medium	Medium
2	None	Low	–	–	–	–	–	–	–	–	–	–
3	–	–	–	–	–	–	None	None	–	–	–	–
4	–	–	–	–	–	–	–	–	Medium	None	–	–
5	Low	Low	–	–	–	–	–	–	–	–	–	–
6	–	–	High	Medium	–	–	–	–	–	–	High	Medium
7	High	High	–	–	–	–	–	–	–	–	–	–
8	High	High	–	–	–	–	–	–	–	–	–	–
9	Medium	Medium	–	–	–	–	–	–	–	–	–	–
10	High	High	–	–	–	–	–	–	–	–	–	–
11	–	–	None	None	None	None	None	None	–	–	None	None
12	None	None	–	–	–	–	–	–	–	–	–	–
13	Medium	Medium	–	–	–	–	–	–	–	–	–	–
14	None	None	–	–	–	–	–	–	–	–	–	–
15	–	–	–	–	–	–	None	Low	–	–	–	–
16	None	None	–	–	–	–	–	–	–	–	–	–
17	Low	High	–	–	–	–	–	–	–	–	–	–
18	Medium	Medium	–	–	–	–	–	–	–	–	–	–
19	None	None	–	–	–	–	–	–	–	–	–	–
20	High	High	–	–	–	–	–	–	–	–	–	–

Table 5. Sensitivity analysis showing the effect (entropy) of NM physiochemical properties on the NM hazard node and the nodes of biological effects.

	Biological effects
Physiochemical properties	NM Hazard	Inflammation	Cytotoxicity	Pulmonary effects	Neurological effects	CNS effects	Fibrosis	Genotoxicity
Nanoparticle	0.39	0.24	0.37	0.13	0.07	0.14	0.06	0.09
Surface coatings	0.22	0.15	0.25	0.14	0.05	0.09	0.02	0.03
Surface area	0.15	0.06	0.16	0.04	0.08	0.08	0	0.2
Aggregation	0.09	0.1	0.12	0.05	0.02	0.03	0	0
Particle size	0.08	0.03	0.09	0.02	0.03	0.05	0	0.01
Surface charge	0.07	0.05	0.06	0.03	0	0.01	0	0
Shape	0.06	0.04	0.07	0.03	0.01	0.02	0.01	0.02
Surface reactivity	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Dissolution	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00

Figure 7. The probability (%) of the hazard potential of NMs as predicted by the BN.

Figure 8. Effect of surface charge on the prediction of the high category level of hazard potential of NMs TiO₂, SiO₂, Ag, CeO₂, ZnO. The figures give the probabilities in percentages.

Table 6. The main NM physiochemical properties and the biological effects of 5 NMs (i.e., TiO₂, SiO₂, Ag, CeO₂, ZnO) affecting NM hazard node.

	Nanoparticles
	TiO2	SiO2	Ag	CeO2	ZnO
Physiochemicalproperties	• Particle size (0.28)• Shape (0.26)• Surface coatings (0.26)• Surface area (0.22)	• Shape (0.25) • Surface coatings (0.17)• Particle size (0.07)	• Surface coatings (0.53)• Surface charge (0.37)• Surface area (0.28)• Aggregation (0.22)	• Aggregation (0.42) • Surface coatings (0.39)• Shape (0.34)• Surface area (0.26)	• Surface reactivity (0.16)• Particle size (0.05)• Surface area (0.02)
Biological effects	• Cytotoxicity (1.03)• CNS effects (0.6)• Inflammation (0.44)	• Cytotoxicity (1.24) • Inflammation (0.57) • CNS effects (0.57)	• Cytotoxicity (0.85) • Inflammation (0.53) • CNS effects (0.49)	• Cytotoxicity (0.94) • Inflammation (0.48) • Pulmonary effects (0.46)	• Cytotoxicity (0.6) • CNS effects (0.17) • Pulmonary effects (0.17)

OECD. 2007. OECD guidline for the testing of chemicals: Developmental Neurotoxicity Study. Test no. 426, pp. 1–26.

 Google Scholar

Morgan K. 2005. Development of a preliminary framework for informing the risk analysis and risk management of nanoparticles. Risk Anal 25:1621–35.

 PubMed Web of Science ®Google Scholar

Chuang HC, Juan HT, Chang CN, Yan YH, Yuan TH, Wang JS, et al. 2014. Cardiopulmonary toxicity of pulmonary exposure to occupationally relevant zinc oxide nanoparticles. Nanotoxicology 8:593–604.

 PubMed Web of Science ®Google Scholar

Gordon T, Nadziejko C, Schlesinger R, Chi Chen L. 1998. Pulmonary and cardiovascular effects of acute exposure to concentrated ambient particulate matter in rats. Toxicol Lett 96–97:285–8.

 PubMed Web of Science ®Google Scholar

WHO. 2012. Guidance for immunotoxicity risk assessment for chemicals. Harmonization Project Document No. 10:1–312.

 Google Scholar

Hiramatsu K, Aoyama H, Zen Y, Aishima S, Kitagawa S, Nakanuma Y. 2006. Proposal of a new staging and grading system of the liver for primary biliary cirrhosis. Histopathology 49:466–78.

 PubMed Web of Science ®Google Scholar

Ashcroft T, Simpson JM, Timbrell V. 1988. Simple method of estimating severity of pulmonary fibrosis on a numerical scale. J Clin Pathol 41:467–70.

 PubMed Web of Science ®Google Scholar

Ng C-T, Li JJ, Bay B-H, Yung L-YL. 2010. Current studies into the genotoxic effects of nanomaterials. J Nucl Acids 2010:1–12.

 Google Scholar