Advanced imaging and AI reveal smoking-related toxins in placenta samples

Rice University scientists and collaborators at Baylor College of Medicine (BCM) have demonstrated a new method for detecting the presence of dangerous chemicals from tobacco smoke in human placenta with unprecedented speed and precision.

The research team used a combination of light-based imaging techniques and machine learning (ML) algorithms to identify and label polycyclic aromatic hydrocarbons (PAHs) and their derivatives (PACs)—toxic compounds generated through the incomplete combustion of organic materials. Exposure to these chemicals during pregnancy can result in negative health outcomes such as preterm birth, low birth weight and developmental problems.

“Our work addresses a critical challenge in maternal and fetal health by improving our ability to detect harmful compounds like PAHs and PACs in placenta samples,” said Oara Neumann, a Rice research scientist who is the first author on a study published in Proceedings of the National Academy of Sciences.

“The findings reveal that machine-learning-enhanced vibrational spectroscopy can accurately distinguish between placental samples from smokers and nonsmokers.”

The new method was used to analyze the placentas of women who reported smoking during pregnancy and self-reported nonsmokers, confirming that PAHs and PACs were present only in the samples collected from smokers.

The findings offer a critical tool for environmental and health monitoring, enabling the identification and labeling of harmful toxins associated with smoking as well as other sources such as wildfires, conflagrations, Superfund sites and other high-pollution environments and contaminated products.

“Measuring levels of environmental chemicals in the placenta can give us insight into the exposures that both mom and baby experienced during pregnancy,” said Melissa Suter, an assistant professor of obstetrics and gynecology at BCM. “This information can help us understand how these chemicals can affect the pregnancy and the baby’s development and help scientists inform public health measures.”

The research relied on surface-enhanced spectroscopy, a method that uses specially designed nanomaterials to amplify the way that specific light wavelengths interact with targeted compounds.

In this case, the researchers leveraged the special optical properties of gold nanoshells designed in the Nanoengineered Photonics and Plasmonics research group led by Naomi Halas, University Professor and the Stanley C. Moore Professor of Electrical and Computer Engineering at Rice.

“We combined two complementary techniques—surface-enhanced Raman spectroscopy and surface-enhanced infrared absorption—to generate highly detailed vibrational signatures of the molecules in the placental samples,” said Halas, who is the corresponding author on the study.

Halas, together with Peter Nordlander, the Wiess Chair in Physics and Astronomy and professor of electrical and computer engineering and materials science and nanoengineering at Rice, have made significant contributions to plasmonics, the study of light-induced collective oscillations of free electrons on the surface of metallic nanoparticles.

Surface-enhanced spectroscopy leverages plasmonics to make possible the in-depth study of molecular structures with very high resolution at the trace concentrations found in biological and environmental samples.

The integration of ML algorithms—characteristic peak extraction (CaPE) and characteristic peak similarity (CaPSim)—revealed subtle patterns in the data that would otherwise have gone undetected. CaPE identified key chemical signatures from the complex datasets, while CaPSim matched these signals to known PAH chemical signatures. This outcome showcases the transformative impact of computational tools for medical and public health applications.

Ankit Patel, assistant professor of electrical and computer engineering at Rice and assistant professor of neuroscience at BCM, said that ML served to “tune out the ‘noise’ in the data.”

“It’s like the so-called ‘cocktail-party effect,'” Patel said. “Picture a noisy and crowded room with lots of people talking at once. We are able to focus our attention on a particular conversation only by tuning out the rest—in the same way, machine learning is able to parse through the spectral data associated with PAHs and PACs much more effectively than humans can.”

Subsequent experiments validated the research findings, confirming that the new method provides a functional alternative to traditional, more labor- and time-intensive techniques.

Beyond smoking-related exposure, the research could enable monitoring exposure to environmental toxins after natural disasters or industrial accidents, equipping health care providers with a faster and more reliable way to assess risk and potentially improve fetal and maternal health outcomes.

“This new method offers an unprecedented level of detail,” said Bhagavatula Moorthy, the Kurt Randerath MD Endowed Chair and Professor of Pediatrics and Neonatology at BCM.

“This research lays the groundwork for expanding ultrasensitive PAH- and PAC-detection technology in biological fluids such as blood and urine as well as in the environmental monitoring of PAHs, PACs and other hazardous chemicals in air, water and soil, thereby aiding in human risk assessment.”

Other Rice co-authors include computer science doctoral alum Yilong Ju, who developed the ML algorithm, and Andres Sanchez-Alvarado, an electrical and computer engineering Ph.D. student in the Halas research group who was part of the team that conducted the experiments.

Rice University scientists and collaborators at Baylor College of Medicine (BCM) have demonstrated a new method for detecting the presence of dangerous chemicals from tobacco smoke in human placenta with unprecedented speed and precision.

The research team used a combination of light-based imaging techniques and machine learning (ML) algorithms to identify and label polycyclic aromatic hydrocarbons (PAHs) and their derivatives (PACs)—toxic compounds generated through the incomplete combustion of organic materials. Exposure to these chemicals during pregnancy can result in negative health outcomes such as preterm birth, low birth weight and developmental problems.

“Our work addresses a critical challenge in maternal and fetal health by improving our ability to detect harmful compounds like PAHs and PACs in placenta samples,” said Oara Neumann, a Rice research scientist who is the first author on a study published in Proceedings of the National Academy of Sciences.

“The findings reveal that machine-learning-enhanced vibrational spectroscopy can accurately distinguish between placental samples from smokers and nonsmokers.”

The new method was used to analyze the placentas of women who reported smoking during pregnancy and self-reported nonsmokers, confirming that PAHs and PACs were present only in the samples collected from smokers.

The findings offer a critical tool for environmental and health monitoring, enabling the identification and labeling of harmful toxins associated with smoking as well as other sources such as wildfires, conflagrations, Superfund sites and other high-pollution environments and contaminated products.

“Measuring levels of environmental chemicals in the placenta can give us insight into the exposures that both mom and baby experienced during pregnancy,” said Melissa Suter, an assistant professor of obstetrics and gynecology at BCM. “This information can help us understand how these chemicals can affect the pregnancy and the baby’s development and help scientists inform public health measures.”

The research relied on surface-enhanced spectroscopy, a method that uses specially designed nanomaterials to amplify the way that specific light wavelengths interact with targeted compounds.

In this case, the researchers leveraged the special optical properties of gold nanoshells designed in the Nanoengineered Photonics and Plasmonics research group led by Naomi Halas, University Professor and the Stanley C. Moore Professor of Electrical and Computer Engineering at Rice.

“We combined two complementary techniques—surface-enhanced Raman spectroscopy and surface-enhanced infrared absorption—to generate highly detailed vibrational signatures of the molecules in the placental samples,” said Halas, who is the corresponding author on the study.

Halas, together with Peter Nordlander, the Wiess Chair in Physics and Astronomy and professor of electrical and computer engineering and materials science and nanoengineering at Rice, have made significant contributions to plasmonics, the study of light-induced collective oscillations of free electrons on the surface of metallic nanoparticles.

Surface-enhanced spectroscopy leverages plasmonics to make possible the in-depth study of molecular structures with very high resolution at the trace concentrations found in biological and environmental samples.

The integration of ML algorithms—characteristic peak extraction (CaPE) and characteristic peak similarity (CaPSim)—revealed subtle patterns in the data that would otherwise have gone undetected. CaPE identified key chemical signatures from the complex datasets, while CaPSim matched these signals to known PAH chemical signatures. This outcome showcases the transformative impact of computational tools for medical and public health applications.

Ankit Patel, assistant professor of electrical and computer engineering at Rice and assistant professor of neuroscience at BCM, said that ML served to “tune out the ‘noise’ in the data.”

“It’s like the so-called ‘cocktail-party effect,'” Patel said. “Picture a noisy and crowded room with lots of people talking at once. We are able to focus our attention on a particular conversation only by tuning out the rest—in the same way, machine learning is able to parse through the spectral data associated with PAHs and PACs much more effectively than humans can.”

Subsequent experiments validated the research findings, confirming that the new method provides a functional alternative to traditional, more labor- and time-intensive techniques.

Beyond smoking-related exposure, the research could enable monitoring exposure to environmental toxins after natural disasters or industrial accidents, equipping health care providers with a faster and more reliable way to assess risk and potentially improve fetal and maternal health outcomes.

“This new method offers an unprecedented level of detail,” said Bhagavatula Moorthy, the Kurt Randerath MD Endowed Chair and Professor of Pediatrics and Neonatology at BCM.

“This research lays the groundwork for expanding ultrasensitive PAH- and PAC-detection technology in biological fluids such as blood and urine as well as in the environmental monitoring of PAHs, PACs and other hazardous chemicals in air, water and soil, thereby aiding in human risk assessment.”

Other Rice co-authors include computer science doctoral alum Yilong Ju, who developed the ML algorithm, and Andres Sanchez-Alvarado, an electrical and computer engineering Ph.D. student in the Halas research group who was part of the team that conducted the experiments.