DE-risking therapeutic use in the context of lactation by coupling perturbation studies and machine learning - infants and young children are most likely to survive, grow and develop fully when…

The National Institute of Child Health and Human Development awarded the University of Melbourne $562,250 on June 19, 2026, under the Child Health and Human Development Extramural Research program (CFDA 93.865) to develop machine learning tools and perturbation studies that assess the safety and impacts of maternal medications during lactation on nursing infants and milk composition.

The project addresses a critical evidence gap: approximately 50 percent of postpartum women require medications, yet data on safe medication use during lactation remains limited, forcing many lactating mothers to discontinue breastfeeding or delay needed treatments out of concern for infant exposure. Current methods for assessing maternal medication safety during lactation are restricted in scope, and direct clinical trials involving lactating women face ethical constraints. The research will leverage advances in machine learning methods and perturbation datasets—in which cells are treated with compounds and their transcriptomes profiled—to create data-driven tools for predicting biological impacts and distribution of therapeutic compounds in the lactation context. These tools will support drug discovery and de-risking processes.

The award carries a period of performance from June 19, 2026, to March 31, 2031. The place of performance is listed as Hanover, New Hampshire 03755, though the University of Melbourne is located in Parkville, Victoria, Australia.

DE-risking therapeutic use in the context of lactation by coupling perturbation studies and machine learning - infants and young children are most likely to survive, grow and develop fully when breastfed. While approximately 50% of postpartum women require medications, data on the safe use of medications during lactation is often lacking. Thus, lactating mothers who take prescribed medications may stop breastfeeding or cut the duration short due to concerns about infant exposures or impacts on their milk. They may also delay much-needed treatments in order to breastfeed. Either scenario can result in impaired health for the nursing dyad. Additionally, emerging evidence suggests that some maternal medications can impact milk composition, including protein and FAT content, and by extension, infant health. Despite this, current methods for assessing the safety and impacts of maternal medication use during lactation are limited in scope. Direct clinical trials involving lactating women are limited due to strong ethical concerns. Additionally, there are no tools that can be used in the drug discovery process to DE-risk impacts that medications may have on lactation. Coupled with the rise in chronic maternal medical conditions, the lack of data-driven tools around maternal medication use and lactation is a major GAP since millions of women are lactating at any given time in the united states. There have been significant advances in machine learning (ML) methods that predict the impacts of therapeutic compounds on cell function in the context of cancer, autoimmunity, and infectious diseases. There are also large databases that have enabled the development of ML tools for drug repurposing, redesign, and safety assessment. These methods rely predominantly on perturbation datasets, where cells are treated with compounds and their transcriptomes profiled. These datasets serve as training data that can be used to predict the biological impacts and distribution of compounds, as well as aid in the development of novel compounds. There is a clear need to build tools that broaden our understanding of how therapeutic compounds impact lactation and breast milk composition in order to support healthy decision-making and drug development for breastfeeding mothers. However, there are no models that can do this in the context of lactation, which is the result of a lack of available datasets to build these models. The overarching hypothesis of our work is that perturbation datasets with human lactocytes can be coupled with novel ML approaches to DE-risk medicine use and drug development for lactating women. We will generate perturbation datasets with human cells that make milk (primary human lactocytes) (aim 1). We will then use several 'omics measures to determine what lactational processes and pathways are impacted by FDA-approved compounds and medications commonly used by lactating women. We will then develop novel ML methods to predict how compounds impact lactocytes to aid in the development of safer medicines for lactating women, while also factoring in predictions of drug transport into breastmilk (aim 2). Finally, we will create engineered cell lines to assess the impact of compounds on processes fundamental to lactation, including protein production, in a scalable way (aim 3). Our proposed work will result in a suit of methods to DE-risk and inform in a data-drive way the use of therapeutics by lactating women.