Close the left navigation

The NIEHS Superfund Research Program (SRP) hosted their Risk e-Learning webinar series focused on using artificial intelligence (AI) and machine learning to advance environmental health research. The series featured SRP-funded researchers, collaborators, and other subject-matter experts who aim to better understand and address environmental health issues by applying AI and machine learning approaches to complex issues.

Recent advances in AI and machine learning methods show promise to improve the accuracy and efficiency of environmental health research. Over the course of three sessions, presenters discussed how they use AI and machine learning approaches to improve chemical analysis, characterize chemical risk, understand microbial ecosystems, develop technologies for contaminant removal, and more.


Session I — AI & ML Applications to Understand Chemical Mixtures, Properties, and Exposures and their Relationship to Human Health 
Monday, November 4, 2024, 2:00 - 4:00 PM ET
To view the webinar archive, visit EPA's CLU-IN Training and Events webpage.

The first session featured three presentations discussing the application of machine learning and artificial intelligence techniques to understand chemical exposures and effects on human health.

Naomi Halas, Ph.D., and Ankit Patel, Ph.D., shared updates on their work combining surface-enhanced spectroscopies (Raman and Infrared Absorption) with machine learning algorithms with the goal of developing simple and ultimately low-cost methods for the detection and identification of environmental toxins. As part of their discussion, they shared several approaches, including the use of machine learning algorithms to detect individual constituents in complex mixtures and the use of facial recognition strategies to identify specific chemical toxins in human placenta.

Jacob Kvasnicka, Ph.D., presented on a project he supported while he was a postdoctoral researcher at Texas A&M University SRP Center’s Risk and Geospatial Sciences Core. There, his work involved developing an ML framework for predicting safe exposure levels to chemicals to avoid cancerous and reproductive/developmental effects. Most chemicals lack toxicity data related to human health, and this study uses ML to fill this gap, greatly expanding the ability to characterize chemical risks and impacts.

Trey Saddler gave attendees an overview of ToxPipe – a platform for performing retrieval augmented generation (RAG) over toxicological data. Comprised of a web interface, agentic workflows, and connections to various data sources, ToxPipe enables toxicologists to explore diverse datasets and generate toxicological narratives for a wide range of compounds.

Presenters:

  • Naomi J. Halas, D.Sc., Ph.D., and Ankit Patel, Ph.D., Rice University
  • Jacob Kvasnicka, Ph.D., U.S. Environmental Protection Agency
  • Trey Saddler, M.S., NIEHS, Division of Translational Toxicology
  • Moderator: David M. Reif, Ph.D., NIEHS, Division of Translational Toxicology


Session II — ML & AI Applications to Environmental Engineering & Bioremediation
Wednesday, November 20, 2024, 2:00 - 4:00 PM ET
To view the webinar archive, visit EPA's CLU-IN Training and Events webpage.

The second session featured three speakers discussing how they apply machine learning and artificial intelligence to environmental engineering applications including detecting contaminants and cleaning up the environment using biosensors, microbiome compositions, and screening tools.

Kei-Hoi Cheung, Ph.D., has an extensive history in data science, and has leveraged that expertise to lead natural language processing (NLP) projects in annotating, extracting, and retrieving environmental exposure data. He presented on the use of these NLP methods combined with ontologies in the in the context of scientific literature on emerging water contaminants.

Mohammad Soheilypour, Ph.D., discussed the application of a suite of computational methods to identify and predict microbial metabolism of various chemical compounds, with a focus on gut and environmental microbiomes. He covered the potential application of machine learning models in this context and their integration with other computational methods to enhance both accuracy and utility.

Paul Westerhoff, Ph.D., highlighted the work of his research team utilizing and comparing two advanced multiple data imputation techniques, AMELIA and MICE algorithms, to fill gaps in sparse groundwater quality datasets to support State agencies in prioritizing future sampling activities. Historic water quality databases are often sparse due to financial budgets for collection and analysis, posing challenges in evaluating exposure or water treatment effectiveness – and this project aims to account for those by accurately assessing and managing risks associated with inorganic pollutants using this technology.

Presenters:

  • Kei-Hoi Cheung, Ph.D., Yale University School of Medicine
  • Mohammad Soheilypour, Ph.D., Nexilico Inc.
  • Paul Westerhoff, Ph.D., Arizona State University
  • Moderator: Heather Henry, NIEHS, Superfund Research Program


Session III — ML & AI Applications to Understand Omics, Metabolomics, & Immunotoxicity and Optimize Bioengineering Using Datasets, Models, and Mass Spectrometry
Friday, November 22, 2024, 12:00 - 2:00 PM ET
To view the webinar archive, visit EPA's CLU-IN Training and Events webpage.

The third session featured four speakers discussing how they apply machine learning and artificial intelligence tools to analyze mass spectrometry and microscopy data and optimize models for understanding metabolomics, metabolite pathways, and immunotoxicology.

Grace Peng, Ph.D., is a co-coordinator of the National Institutes of Health (NIH) Common Fund’s Bridge to Artificial Intelligence (Bridge2AI) program, bridging the gap between the biomedical, behavioral and bioethics research communities and the data science/AI communities through a consortium of diverse experts to set the stage for widespread adoption of AI/ML in medicine. Dr. Peng gave an overview of the Bridge2AI program and introduce one of their projects at the University of California San Diego – Trey Ideker, Ph.D. Dr. Ideker discussed the cell maps for AI (CM4AI) functional genomics project, one of four major data generation projects under the Bridge2AI program. The goal of the project is to provide a comprehensive map of human cellular components through generation of major spatial proteomics datasets.

John Efromson, M.S., presented on Ramona Optic, Inc.’s Multi-Camera Array Microscope [MCAM(TM)], which is used to automate imaging and computer vision analysis of zebrafish and greatly improves previous throughput and analysis capabilities. Multiple applications of machine learning were discussed, including behavioral pose estimation and phenotyping, morphological analysis, and cell counting and fluorescence quantification, as well as how these distinct analyses can be used together for pharmacology, toxicology, and neuroscience research.

Forest White, Ph.D., spoke on the innovations being pursued by the MIT Superfund Research Center’s Data Management and Analysis Core (DMAC). The MIT DMAC has been working on a cutting-edge mass spectrometry approach for protein phosphorylation profiling, next-generation sequencing for transcript expression profiling, and computational modeling to integrate molecular network data with cell phenotypic data.

Presenters:

  • Grace C.Y. Peng, Ph.D., Division of Discovery Science and Technology (Bioengineering), NIBIB and Trey Ideker, Ph.D., University of California San Diego
  • John Efromson, M.S., Ramona Optics
  • Forest White, Ph.D., MIT Department of Biological Engineering
  • Moderator: Hunter Moseley, Ph.D., University of Kentucky