Walther-António, Marina

Abstract Disclosure: A. Zhang: None. J. Kapoor: None. C. Deveney: None. A.G. Cope: None. M.M. Walther-Antonio: None. N. Chia: None. C.W. Pattanayak: None. M.J. Tetel: None. The vaginal microbiome (VM) consists of the microorganisms and their metabolites residing in the vaginal tract and is critical for maintaining female reproductive health. The Lactobacillus species predominate most VMs and produce lactic acid, which has been cited as a key protector of vaginal health. Alterations in the finely balanced vaginal microbiome can cause infections, such as bacterial vaginosis, and may contribute to the development of endometrial and ovarian cancer. While it is well-established that stress and the gut microbiome impact each other via the gut-brain axis, the influence of stress, hormones, and lifestyle factors on the VM composition remains poorly understood. We investigated the relationship between stress, hormones (menstruation, hormonal contraception), lifestyle factors (diet, exercise, sleep), and the VM composition. In this 10-week study, healthy, young female volunteers (n=62) provided daily vaginal and fecal swabs, weekly stress ratings, daily data on sleep and activity (via FitBit) and diet information. High-throughput 16S rRNA sequencing was used to identify the vaginal microbial composition of each sample. Types of hormonal birth control were none (n=30), local progestin (n=9), systemic progestin-only (n=7), or systemic estrogens and progestin combined (n=16). Consistent with our previous study, vaginal microbial diversity, as measured by the Shannon index, varied by type of hormonal birth control, with participants not on hormonal birth control having the highest average diversity. There was a positive relationship between average stress and vaginal microbial diversity, with preliminary results suggesting that this relationship depends on hormonal birth control status. Participants who menstruated during the 10-week study had a higher vaginal microbial diversity than those who did not. We are currently analyzing which microbes contribute to these differences in vaginal microbial diversity. Microbiota and fungal data from the daily vaginal and fecal swab samples will be analyzed to determine how stress, hormones and/or lifestyle factors relate to crosstalk between the vaginal and gut microbiomes. These findings may provide long-term, time-longitudinal insights into how these two important microbiomes are involved in important aspects of women’s health. Presentation: Sunday, July 13, 2025

Abstract In this work, we seek to characterize the learned feature space of cancer drug response prediction models in a drug-blind setting and quantify the limits of their generalizability. Drug-blind prediction failure describes the inability of models to predict cell line response to drugs in the test set that are not present during training. The experiments performed in this study utilize a two-arm multilayer perceptron model where embeddings for drugs and cell-lines are calculated separately and then concatenated to predict response. Drug structure is represented by Morgan fingerprint while cell lines utilize gene expression values. We first examine the connection between learned drug features and cell lines by permuting responses within both cell lines and drugs during model training. When permuting response values within cell lines, model performance was entirely depleted, but permutation of responses within drugs resulted in only a 10-15% decrease in performance. This displays the bulk of model performance is due to learned distributions of response for each drug. We also determine the impact of dataset size on drug-blind performance. Increasing the cell line examples per drug did not improve performance, but drug-blind performance exhibited higher variance than mixed set testing. From these experiments, we hypothesized that drug-blind performance was a function of the set of drugs in the training set. We trained a set of 208 models with varying training set and constant test set and fit an elastic net model using drugs in the training set as features and performance of each model as the target. We were able to accurately predict the drug-blind performance of a model, measured in Pearson correlation, based on the drugs in the training set with a mean absolute error of 0.049. We then trained a set of 1641 models where all drugs in the dataset were present in the test set but only 50% of unique drugs were present during training. Hierarchical clustering on the coefficients of an elastic net model fit to each unique drug’s performance across all models shows drugs cluster into groups that correspond to specific mechanisms of action. By examining the relationship between drug embeddings during model training, we see that decreasing validation loss corresponds to reinforcement of mechanistic relationships of drugs not captured by raw Morgan fingerprints. Finally, we show that training on a dataset confined to a single mechanism of action significantly improves overall mixed set performance on those drugs against training on the entire set of drugs. In this study, we identify that drug-blind performance in current large pharmacogenomic datasets is limited to generalization within drug mechanism of action classes. Therefore, global drug-blind prediction benchmarking is a poor indicator of model generalization as the data itself creates these limits, not model architecture. As more data is collected on novel anti-cancer compounds, we hope that the results presented here create a foundation on which to measure progress in cancer drug generalization. Citation Format: William G. Herbert, Paul A. Jensen, Nicholas Chia, Marina RS. Walther-Antonio. Monotherapy cancer drug-blind response prediction is limited to intraclass generalization [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Artificial Intelligence and Machine Learning; 2025 Jul 10-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(13_Suppl):Abstract nr A005.

Abstract Breast cancer is the leading cause of cancer death among women worldwide. Most cases occur in those without identifiable modifiable risk factors, highlighting critical gaps in our understanding of breast cancer biology including the influence of the gut microbiome. Here we studied the gut microbiome with shallow shotgun metagenomic sequencing from preoperative stool samples in a well-characterized clinical cohort of 131 women, median age 59 years, undergoing breast-conserving surgery for benign (n=23), high-risk (n=47), or malignant (n=61) breast disease. Taxonomic profiling was performed with Sourmasth 4.2.4 with GTDBv202 as reference database; functional profiling of metagenomic data employed HUMAnN 3.6 with identified gene families mapped to MetaCyc pathways. Differences in alpha diversity in terms of species richness were observed across diagnosis groups (benign, high-risk, malignant) and was lowest for high-risk lesions (p=0.031). Decreased alpha diversity was also associated with higher body mass index (BMI) (Shannon p=0.12, Inverse Simpson p=0.025). Significant beta diversity differences across diagnosis groups also were seen (Bray-Curtis p=0.044, Jaccard p=0.013) as were differences in association with BMI (Bray-Curtis, Jaccard, both p=0.002). No significant diversity differences associated with age, mammographic breast density or menopausal status were observed. Functional pathway profiling stratified by BMI showed a correlation between increasing BMI and greater relative abundance of genes in nicotinamide adenine dinucleotide (NAD) biosynthesis, folate metabolism, and aromatic amino acid metabolism pathways (p=0.0279, 0.0667 and 0.0860, respectively) - pathways implicated in breast cancer detection, survival and prognosis. Cross-referencing our significantly (FDR <5%) differentially abundant taxonomic findings with these functional pathway data identified enrichment of several Blautia species, linked to NAD biosynthesis via tryptophan metabolism, in patients with higher BMI. We also noted enrichment of Dorea formicigenerans, a species associated with increased breast cancer risk in obese women due to its pro-inflammatory effects, in individuals with a higher BMI. Here we report significant differences in gut microbiome diversity, taxonomy and functional pathway signals in association with BMI in women with breast disease. Differentially abundant species seen with rising BMI may contribute to functional pathways that drive increased breast cancer risk and may provide an actionable target for risk reduction independent of conventional risk reduction strategies. Citation Format: Jack W. Sample, Matteo Redaelli, Stephen Johnson, Tanya L. Hoskin, Marina R. Walther-Antonio, Jun Chen, Tina J. Hieken. Obesity alters cancer-related microbial pathways in women with benign and malignant breast disease [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2213.

Objectives/Goals: Longitudinal studies capture dynamic host–environment interactions crucial to microbiome research. Commonly used mixed-effects models can struggle with small clinical samples, like many endometrial cancer studies. We will compare them to simpler fixed-effects models, aiming to develop a new fixed-effects differential abundance alternative. Methods/Study Population: Starting with simulated data for linear models with a single treatment effect, we will incrementally increase complexity to reflect real-world microbiome data. To observe conditions of misspecification, simulated data of repeated measures will be generated using different distributions, while varying parameters and inter-class correlation. In each condition, linear fixed- and mixed-effects models will be fitted 1000 times to empirically determine p-value, type 1 error rate, and power. We will increase the verisimilitude of our approach by using endometrial cancer data from our lab as a real background, while adding simulated signals. This data will be used to assess extant differential abundance models against our novel fixed-effects model. Results/Anticipated Results: Our initial findings indicate that fixed-effects models maintain control over type 1 error rates in small samples, while mixed-effects models do not perform as reliably. We are now exploring the effects of confounders on type 1 error rates and power in linear models, while our next step will evaluate generalized linear models in a differential abundance context. We expect that the fixed-effects models will continue to be as reliable as mixed-effects models, while being less computationally complex. After establishing our theoretical basis, our novel model should perform comparably to mixed-effects models on pseudo-simulated endometrial cancer data and better at small sample sizes. We will finally apply our model to real endometrial cancer data to identify microbial markers of disease predisposition. Discussion/Significance of Impact: By characterizing and improving these statistical tools, we seek to increase the robustness and power of microbiome study results (especially at small sample sizes), ultimately supporting better-informed translational research decisions. This work underscores the importance of statistical methods in advancing microbiome research.

We will do a recap of our microbiome work in Obstetrics and Gynecology, starting with our findings in healthy populations and natural variation. Then transitioning to healthy obstetrics and preterm labor. And wrapping up with gynecologic cancers, particularly endometrial and ovarian cancer.

Motility is a great biosignature and its pattern is characteristic for specific microbes. However, motion does also occur within the cell by the myriads of ongoing processes within the cell and the exchange of gases and nutrients with the outside environment. Here, we propose that the sum of these processes in a microbial cell is equivalent to a pulse in complex organisms and suggest a first approach to measure the “living pulse” in microorganisms. We emphasize that if a “living pulse” can be shown to exist, it would have far-reaching applications, such as for finding life in extreme environments on Earth and in extraterrestrial locations, as well as making sure that life is not present where it should not be, such as during medical procedures and in the food processing industry.

Background: Primary sclerosing cholangitis (PSC) is a major risk factor for cholangiocarcinoma (CCA). We investigated biliary and fecal microbiota to determine whether specific microbes in the bile or stool are associated with PSC or CCA. Methods: Bile was obtained from 32 patients with PSC, 23 with CCA with PSC, 26 with CCA without PSC, and 17 controls. Over 90% of bile samples were from patients with perihilar CCA. Stool was obtained from 31 patients with PSC (11 were matched to bile), 16 with CCA with PSC (10 matched to bile), and 11 with CCA without PSC (6 matched to bile). Microbiota composition was assessed using 16SrRNA-marker-based sequencing and was compared between groups. Results: Bile has a unique microbiota distinguished from negative DNA controls and stool. Increased species richness and abundance of Fusobacteria correlated with duration of PSC and characterized the biliary microbiota in CCA. Stool microbiota composition showed no significant differences between groups. Conclusions: We identified a unique microbial signature in the bile of patients with increased duration of PSC or with CCA, suggesting a role for microbiota-driven inflammation in the pathogenesis and or progression to perihilar CCA. Further studies are needed to test this hypothesis.

Abstract Whole genome sequencing is emerging as a promising tool for the untargeted detection of a broad range of microbial species for diagnosis and analysis. However, it is logistically challenging to perform the multistep process from sample preparation to DNA amplification to sequencing and analysis within a short turnaround time. To address this challenge, we developed a digital microfluidic device for rapid whole genome amplification of low-abundance bacterial DNA and compared results with conventional in-tube DNA amplification. In this work, we chose Corynebacterium glutamicum DNA as a bacterial target for method development and optimization, as it is not a common contaminant. Sequencing was performed in a hand-held Oxford Nanopore Technologies MinION sequencer. Our results show that using an in-tube amplification approach, at least 1 pg starting DNA is needed to reach the amount required for successful sequencing within 2 h. While using a digital microfluidic device, it is possible to amplify as low as 10 fg of C. glutamicum DNA (equivalent to the amount of DNA within a single bacterial cell) within 2 h and to identify the target bacterium within 30 min of MinION sequencing100× lower than the detection limit of an in-tube amplification approach. We demonstrate the detection of C. glutamicum DNA in a mock community DNA sample and characterize the limit of bacterial detection in the presence of human cells. This approach can be used to identify microbes with minute amounts of genetic material in samples depleted of human cells within 3 h.

Abstract Background As the use of antibiotics during the peripartum period increases, the incidence of autoimmune disorders and autism spectrum disorders (ASDs) is also increasing. In this study, we aim to assess if antibiotic exposure during the peripartum period affects the incidence of autoimmune diseases and ASD in the offspring. Methods We identified children (< 18 years of age) born in Olmsted County from January 1, 2003 through December 31, 2012. Offspring with celiac disease (CD), inflammatory bowel disease (IBD), or ASD diagnoses were matched to two controls on birth date, index date, mother’s age at delivery, and sex. Data from the mother’s medical records were retrieved to determine peripartum antibiotics use. Results A total of 242 cases and 484 matched controls were included in this study. Median age at the last follow-up was 11.3 years (range: 0.5–14.9), 73% were males in both groups. Odds of CD diagnosis was not statistically different between vaginal delivery with antibiotics compared with vaginal delivery with no antibiotics (odds ratio [OR] = 0.76, 95% confidence interval [CI]: 0.32–1.85), similarly in IBD (OR = 2.41, 95% CI: 0.53–10.98) and ASD (OR = 1.00, 95% CI:0.55–1.79). Preeclampsia or eclampsia was associated with offspring CD (OR = 3.20, 95% CI: 1.05–9.78). Smoking history and diabetes mellitus were associated with offspring ASD (OR = 1.84, 95% CI: 1.22–2.77 and OR = 2.01, 95% CI: 1.03–3.91, respectively). Conclusion In this cohort, we found no statistically significant association between peripartum antibiotics exposure and the development of CD, IBD, or ASD.