Keaton Smith

sciencedirect.comMachine learning in placental research: a review and guide for future studiesMachine learning approaches are increasingly applied across biomedical research to analyze complex, high-dimensional datasets and identify patterns th…

Last week we published a review paper in Placenta focused in the use of machine learning (ML) in placental multi-omics research. www.sciencedirect.com/science/arti...

ML models are only as reliable as the data and study designs behind them. Poor data quality, technical artifacts, and unrepresentative cohorts can produce biased models. Our review outlines best practices to address limitations in placental ML studies and improve model robustness and reliability.

Finally, we highlight the need for open access to high-quality placental multi-omics datasets. We encourage depositing data in public repositories with detailed metadata to enable independent validation, cross-cohort comparisons, and benchmarking of different analytical approaches.

Last week we published a review paper in Placenta focused in the use of machine learning (ML) in placental multi-omics research. www.sciencedirect.com/science/arti...

In this review, we discuss key ML concepts, emerging applications across placental biology, and the methodological challenges that need to be addressed to make these approaches more robust, reproducible, and clinically meaningful.

We explore emerging machine learning applications in placental research across four main areas: genomics, epigenomics, transcriptomics, and proteomics. ML approaches have been applied to investigate placental biology, pregnancy complications, fetal development, and early childhood outcomes.

doi.orgEnrichment of methylated cell-free placental DNAIntroduction Preterm birth drives adverse perinatal maternal and infant health outcomes through heterogeneous symptoms, severity, and etiologies. Delivery prior to reaching 37 weeks of gestation may result from medically indicated intervention for pregnancy complications or spontaneously in the absence of prior symptoms. Placental tissue collected following preterm birth exhibits differential DNA methylation compared to full-term placentas and may indicate pregnancy health during gestation. Placental DNA currently has limited utility for assessing health of ongoing pregnancy, as sampling placental tissue during gestation increases the risk of infection and miscarriage. Risks associated with placental sampling during pregnancy limit the use of DNA methylation in clinical preterm birth prediction. Assessing preterm birth risk during gestation requires non-invasive methods for characterizing placental DNA methylation. Results We quantified genome-wide DNA methylation patterns of hypermethylated cell-free DNA in pregnant ( n = 99) and non-pregnant ( n = 93) plasma using cell-free methylated DNA immunoprecipitation sequencing (cfMeDIP-seq). In each sample, we assessed DNA methylation status in 300-bp genomic windows, examining both sequencing read counts and calculated absolute molar DNA amount. Known hypermethylated placental regions, including RASSF1 , STAT5A , and ERG promoters showed significantly increased odds of detection in pregnant samples, suggesting en-richment of cell-free placental DNA. Of the 536,444 300-bp windows examined, 173,071 (32%) showed significant enrichment in pregnant plasma. Linear modeling identified 107,505 differentially methylated regions (DMRs) associated with pregnancies later diagnosed with intrauterine growth restriction (IUGR) ( n = 22). Alu elements showed increased representation in these DMRs than expected, while other repetitive elements exhibited underrepresentation. Discussion These results demonstrate cfMeDIP-seq’s ability to enrich for cell-free placental DNA and characterize cell-free DNA methylation signatures of pregnancies complicated by IUGR. Enrichment of cell-free placental DNA enables non-invasive profiling of placental DNA methylation from maternal plasma. Detectable epigenetic signatures in maternal plasma may identify pregnancies at elevated risk for preterm birth before clinical symptoms appear. Our findings further highlight the potential of cell-free placental DNA for monitoring pregnancy health. ### Competing Interest Statement S.Y.S., T.T., D.D.De C., S.L.W., and M.M.H. are inventors on a patent related to the spike-in controls, licensed to Adela. S.Y.S., S.V.B., and D.D.De C. are inventors on other patent applications related to cell-free DNA methylation analysis technologies, licensed to Adela, serve in leadership roles at Adela, and own equity in Adela. S.V.B. is inventor on a patent related to cell-free DNA mutation analysis technologies, licensed to Roche Molecular Diagnostics. S.V.B. and D.D.De C. have received research funding from Nektar Therapeutics. Canadian Institutes of Health Research, 389866 McLaughlin Centre – University of Toronto, MC-2019-07

In August, we put out a preprint exploring whether cell-free DNA methylation in maternal blood could provide a non-invasive way to assess placental health during pregnancy. doi.org/10.64898/202...

We identified a strong cell-free DNA methylation signature in pregnancies later diagnosed with intrauterine growth restriction. This suggests that cell-free placental DNA methylation patterns may indicate DNA methylation alterations associated with impaired fetal growth prior to symptom development.

These results support cell-free DNA methylation profiling as a non-invasive approach to monitoring placental health during pregnancy. Further work is needed to validate these signatures and determine whether they can reliably predict pregnancy complications and inform clinical care.

In August, we put out a preprint exploring whether cell-free DNA methylation in maternal blood could provide a non-invasive way to assess placental health during pregnancy. doi.org/10.64898/202...

We used cfMeDIP-seq to profile genome-wide DNA methylation in plasma from 99 pregnancy and 93 non-pregnant individuals. We found enrichment of placental cell-free DNA by cfMeDIP-seq, where regions with known hypermethylation in placenta were more likely to be detected in pregnant participants.

doi.orgDistinct DNA methylation signatures in maternal blood reveal unique immune cell shifts in preeclampsia and the pregnancy-postpartum transitionPreeclampsia (PE) is a hypertensive disorder of pregnancy characterized by immune dysregulation and significant risks to maternal and fetal health. While current management relies on high-risk patient monitoring and early diagnosis, these methods are costly and burdensome, especially for low-risk pregnancies. DNA methylation (DNAm) is a type of chemical modification that influences gene expression and has been associated with immune cell dynamics and PE pathogenesis. This study explores whether DNAm-based immune cell composition profiling can provide insights into immune dysregulation associated with PE. By also examining changes in immune cell composition across gestational timepoints and into the postpartum period, we aimed to establish a baseline of healthy immune adaptation during pregnancy, against which PE-related disruptions can be better understood. We conducted a search in the Gene Expression Omnibus (GEO) for DNAm datasets using Illumina 27K, 450K, and EPIC arrays from maternal blood in both healthy and PE pregnancies. We found two studies (GSE37722 and GSE192918) that met our criteria, involving a total of 24 healthy pregnancies and 14 with PE. To estimate immune cell composition (CD8 + T cells, CD4 + T cells, monocytes, granulocytes, natural killer cells, and B cells) from DNAm data, we applied the deconvolution algorithm developed by Houseman et al (2012). A linear model was used to assess statistical differences in immune cell proportions between PE cases and controls. Longitudinal analyses were also conducted to examine immune cell shifts during pregnancy and postpartum. No significant differences were observed between PE and control groups in any immune cell type. However, longitudinal analyses revealed substantial immune remodeling in the postpartum period, characterized by decreased monocytes and granulocytes, and increased natural killer cells, B cells, and T cells. While subgroup analyses showed some variability in significance, particularly in GSE192918, the overall trends were consistent across datasets, emphasizing the importance of gestational age in immune dynamics. These findings support the use of DNAm profiling as a valuable tool for characterizing immune cell dynamics during pregnancy. Although immune differences between PE cases and controls were not observed with the Houseman method, longitudinal shifts were consistently captured and provide additional insights into the evolution of immune changes from pregnancy to postpartum, supporting the potential of DNAm-based profiling for developing predictive and monitoring tools for pregnancy and pregnancy-related pathology. It is important to note that these analyses were based on a single deconvolution approach applied to a cohort with well-matched clinical criteria; and that differences in study design, timing of sample collection, and cohort characteristics may limit broader generalizability. Future studies leveraging pregnancy-included reference matrices in deconvolution methods and larger, more diverse cohorts are essential to refine the application of DNAm-based immune profiling in pregnancy and pregnancy complications.

It's been a busy year, so I'm excited to share some of my recent work from @samanthalwilson.bsky.social lab! In February, I co-authored a PLOS One paper with Dr. Laiba Jamshed, demonstrating DNA methylation-based profiling to characterize immune cell dynamics during pregnancy doi.org/10.1371/jour...

Although we didn't detect overall differences in immune cell proportions when comparing preeclampsia and healthy pregnancies, we identified immune remodeling after delivery. Our findings support the potential use of DNA methylation-based monitoring tools for pregnancy and related pathology.

Keaton Smith reposted

Melanie Lemaire

@m-lemaire.bsky.social

26 Sept 2025

Very excited to share the results from our latest Meta-analysis at the Wilson Pregnancy Lab examining #placental DNA methylation in #IVF pregnancies doi.org/10.1093/mole... @smithk63.bsky.social @samanthalwilson.bsky.social @oxfordacademic.bsky.social

13 Dec 2024

Of course our cats waited until I was on a Zoom call before tearing around the apartment and having a loud wrestling match right next to my desk #catdad

journals.plos.orgCell-free placental DNA: What do we really know?Author summary Cell-free DNA are fragments of free-floating DNA released by cells into blood circulation. During pregnancy, the placenta releases cell-free placental DNA (cfpDNA) into maternal blood c...

10 Dec 2024

Congratulations to my lab mates for a phenomenal paper from @samanthalwilson.bsky.social lab: journals.plos.org/plosgenetics...

14 Nov 2024

For my colleagues in placental research, the November installment of the North American Placenta Lab Meeting is happening this Tuesday (Nov 19) at 9:00 am PST/12:00 pm EST over Zoom (mcmaster.zoom.us/j/96005378701). See details on our scheduled speakers below!

I've had a Bluesky account for a little while, but I'm hoping to start getting more use out of it now! I thought I would do a few posts to introduce myself, my experience, and my interests. So hi Bluesky community, I'm Keaton!

forms.gleNorth American Placenta Lab Meeting - mailing list sign-upSign up below to receive Zoom links to the monthly Canadian Placenta Lab Meeting Please email Keaton Smith (smithk63@mcmaster.ca) for Zoom details if you are interested in joining the January 2024 we...

These days, I also help organize the North American Placenta Lab Meeting, a monthly virtual seminar series where placenta researchers can showcase their work to the community (forms.gle/27VExgd2bQCJ...).

I'm excited to hopefully connect with fellow grad students and other researchers here, as well as keep up to date on research from my favourite labs and colleagues. Thanks for reading if you've made it this far!

wilsonpregnancylab.comWilson Pregnancy LabUndergraduate Thesis Student - 2023

Looking to venture into human genetics and expand my computational skill set, I was lucky to land a co-op and thesis position with Dr. Samantha Wilson (www.wilsonpregnancylab.com). My project looked at the impact of maternal smoking on cell-free placental DNA methylation.

I'm now a graduate student with the lab examining placental multi-omic responses to environmental exposures developing non-invasive models to predict pregnancy complications such as preeclampsia and intrauterine growth restriction.

doi.orgA standardized nomenclature for resistance-modifying agents in the Comprehensive Antibiotic Resistance Database | Microbiology SpectrumWhile increasing rates of antimicrobial resistance undermine our current arsenal of antibiotics, resistance-modifying agents (RMAs) hold promise to extend the lifetime of these important molecules. We...

I also managed to land my own short paper for overhauling and standardizing CARD's curation practices and nomenclature for resistance-mofidying agents (doi.org/10.1128/spectrum.02744-23).

I learned so much about bioinformatics and had such a good time that when my placement officially ended, I stayed on to lead a team of volunteers formalizing glycopeptide resistance cluster curation.

3 years, 2 co-op placements, and a thesis later, I graduated from the BioPharm program with an Honours Bachelor of Health Science.

card.mcmaster.caThe Comprehensive Antibiotic Resistance Database

My first placement was in the lab of Dr. Andrew McArthur as a junior curator for the Comprehensive Antibiotic Resistance Database (CARD, card.mcmaster.ca). I was mainly responsible for systematic curation, which earned me co-authorship on the 2023 CARD update paper (doi.org/10.1093/nar/gkac920).