Evaluation of carcinogenic modes of action for pesticides in fruit on the Swedish market using a text-mining tool

Ilona Silins, Anna Korhonen, Ulla SteniusFrontiers in Pharmacology 2014 · DOI

Abstract

Toxicity caused by chemical mixtures has emerged as a significant challenge for toxicologists and risk assessors. Information on individual chemicals' modes of action is an important part of the hazard identification step. In this study, an automatic text mining-based tool was employed as a method to identify the carcinogenic modes of action of pesticides frequently found in fruit on the Swedish market. The current available scientific literature on the 26 most common pesticides found in apples and oranges was evaluated. The literature was classified according to a taxonomy that specifies the main type of scientific evidence used for determining carcinogenic properties of chemicals. The publication profiles of many pesticides were similar, containing evidence for both genotoxic and non-genotoxic modes of action, including effects such as oxidative stress, chromosomal changes and cell proliferation. We also found that 18 of the 26 pesticides studied here had previously caused tumors in at least one animal species, findings which support the mode of action data. This study shows how a text-mining tool could be used to identify carcinogenic modes of action for a group of chemicals in large quantities of text. This strategy could support the risk assessment process of chemical mixtures.

Introduction: The Mixture Problem in Toxicology

Toxicologists currently face a daunting "mixture problem." While regulatory frameworks typically assess and regulate pesticides individually, real-world exposure—particularly through diet—often involves a cocktail of different chemical residues. To understand if these chemicals act together, either additively or synergistically, risk assessors require detailed knowledge of their Mode of Action (MOA). An MOA describes the specific biological sequence of events that leads to a toxic effect, such as cancer.

However, identifying MOAs is hindered by the sheer volume of scientific literature. Manually reviewing thousands of papers for dozens of chemicals is a slow, labor-intensive process that creates a bottleneck in hazard identification. This study addresses this challenge by applying automated tools to accelerate the identification of carcinogenic MOAs for pesticides frequently detected in fruit on the Swedish market.

Methodology: Automated Hazard Identification

The researchers utilized the CRAB (Cancer Risk Assessment tool), a specialized text-mining system designed for cancer risk assessment. The methodology followed a structured pipeline:

  1. Taxonomy-Based Classification: CRAB utilizes a taxonomy of 25 sub-categories (e.g., oxidative stress, cell proliferation, hormonal effects) nested under two primary umbrellas: genotoxic (direct DNA damage) and non-genotoxic (cancer promotion via other biological pathways).
  2. Automated Sifting: The tool automatically retrieved and analyzed 24,424 PubMed abstracts related to 26 pesticides identified by the Swedish National Food Agency (NFA) as common residues in apples and oranges.
  3. Profiling: The system generated "publication profiles" for each chemical, representing the percentage of literature associated with specific biological effects.
Figure 1: Schematic flow chart of the CRAB tool's literature classification process, showing the reduction of over 24,000 references to a relevant subset.

Key Results and Insights

The study demonstrated that text mining can significantly reduce the manual workload of risk assessment. Of the initial 24,424 abstracts, CRAB identified 6,087 (approximately 25%) as truly relevant to carcinogenic MOA analysis.

MOA Publication Profiles

The analysis revealed that many pesticides share remarkably similar publication profiles. Evidence for oxidative stress and chromosomal changes was prevalent across the studied group, suggesting that these chemicals might have additive effects when consumed together. The tool also highlighted data gaps; while some pesticides had thousands of associated abstracts, others, such as boscalid, had very little relevant MOA data, indicating a need for further primary research.

Figure 2: Literature distribution profiles across MOA categories for pesticides found in apples (A) and oranges (B), with mean comparisons (C).

Validation and Mixture Prevalence

The tool's findings were validated against established cancer classifications from the US EPA and IARC. Notably, 18 of the 26 pesticides had previously been shown to cause tumors in at least one animal species. The urgency of this research is underscored by the prevalence of mixtures: 78% of apple samples and 96% of orange samples contained multiple pesticide residues.

Discussion and Limitations

While the CRAB tool proved highly efficient, several limitations remain. The current analysis is based primarily on abstracts; the researchers noted that full-text analysis is a necessary next step to extract more granular data. Furthermore, while text mining is excellent for identifying qualitative hazards (what a chemical can do), it struggles to extract the quantitative dose-response data required for final risk characterization.

Ultimately, this study provides the "hazard" side of the risk equation. Future work must integrate these MOA profiles with actual human exposure levels to estimate real-world risk accurately. As our understanding of epigenetics and non-genotoxic pathways evolves, the MOA taxonomy will require continuous updates to remain a robust tool for chemical mixture assessment.

Summary updated

Abstract

Toxicity caused by chemical mixtures has emerged as a significant challenge for toxicologists and risk assessors. Information on individual chemicals' modes of action is an important part of the hazard identification step. In this study, an automatic text mining-based tool was employed as a method to identify the carcinogenic modes of action of pesticides frequently found in fruit on the Swedish market. The current available scientific literature on the 26 most common pesticides found in apples and oranges was evaluated. The literature was classified according to a taxonomy that specifies the main type of scientific evidence used for determining carcinogenic properties of chemicals. The publication profiles of many pesticides were similar, containing evidence for both genotoxic and non-genotoxic modes of action, including effects such as oxidative stress, chromosomal changes and cell proliferation. We also found that 18 of the 26 pesticides studied here had previously caused tumors in at least one animal species, findings which support the mode of action data. This study shows how a text-mining tool could be used to identify carcinogenic modes of action for a group of chemicals in large quantities of text. This strategy could support the risk assessment process of chemical mixtures.
BibTeX
@inproceedings{silins2014evaluation,
  author = {Ilona Silins and Anna Korhonen and Ulla Stenius},
  title = {{Evaluation of carcinogenic modes of action for pesticides in fruit on the Swedish market using a text-mining tool}},
  booktitle = {Frontiers in Pharmacology},
  year = {2014},
  doi = {10.3389/fphar.2014.00145},
}
Figure