Visualization strategies to aid interpretation of high-dimensional genotoxicity data

Area of Science:

• Toxicology
• Data Science
• Computational Biology

Background:

• Multiplexed assays generate high-dimensional toxicological data.
• Machine learning algorithms are increasingly used for analyzing such data.
• Effective visualization is crucial for interpreting complex datasets.

Purpose of the Study:

• To explore and compare high-dimensional data visualization strategies.
• To assess their utility in complementing machine learning predictions from MultiFlow® assay results.
• To aid in the interpretation of toxicological data from chemical exposures.

Main Methods:

• Evaluation of various visualization techniques including scatter plots, spider plots, parallel coordinate plots, hierarchical clustering, principal component analysis, multidimensional scaling, t-distributed stochastic neighbor embedding, and uniform manifold approximation and projection.
• Application to a dataset of seven biomarker responses from TK6 cells exposed to 126 chemicals at various concentrations.
• Comparative analysis of the effectiveness of each visualization method.

Main Results:

• Different visualization techniques offer unique strengths for exploring high-dimensional toxicological data.
• Certain methods effectively represent complex relationships between chemicals, biomarkers, and concentrations.
• Visualization aids in understanding patterns that complement machine learning model outputs.

Conclusions:

• High-dimensional visualization strategies are valuable tools for toxicological data interpretation.
• These methods enhance the understanding of multiplexed assay results and machine learning predictions.
• Effective visualization facilitates efficient and accurate decision-making in chemical safety assessment.