Biology subjects

Glettig, M.

Publications and source records attributed to Glettig, M..

4 recordsLinked to original sources

bioRxiv2025

H&Enium, Applying Foundation Models to Computational Pathology and Spatial Transcriptomics to Learn an Aligned Latent Space

Bridging the gap from transcriptomic to imaging data at single-cell resolution is essential for understanding tumor biology and improving cancer diagnostics. Spatial transcriptomics enables mapping gene expression onto H&E images of segmented single cells, but remains limited by cost and throughput. We introduce H&Enium, a contrastive alignment framework that projects image and gene expression embeddings from foundation models into an aligned latent space using projection heads and a novel soft alignment target. This alignment enriches image-derived embeddings with transcriptomic context improving downstream tasks such as cell type classification and gene expression prediction. Additional evaluations on independent pathology datasets demonstrate superior generalization of our aligned representations over unaligned baselines. Our method offers a scalable path to enhance the utility of standard H&E imaging in both research and clinical settings.

pathology↗

bioRxiv2025

Clinico-genomic features predict distinct metastatic phenotypes in cutaneous melanoma

Metastasis drives mortality and morbidity in cancer. While some patients develop broad metastatic disease across multiple organs, others exhibit organ-specific spread. To identify mechanisms underlying metastatic organotropism, we analyzed clinico-genomic data from over 7,000 patients with metastatic cutaneous melanoma in three independent cohorts (one primary discovery and two validation cohorts including a nationwide electronic health record-derived deidentified database), leveraging machine learning approaches to clinical data. We found that female sex and increased tumor mutational burden associate with decreased metastatic potential, while older age associates with increased lung and adrenal metastases. Using unsupervised analyses, patients clustered into five metastatic patterns: a "highly metastatic" cluster characterized by involvement of many organs, a "low metastatic" cluster characterized by few metastatic sites (mostly lymph node metastases), and three additional clusters each characterized by metastasis to specific sites (brain, lung, liver). Mutations in B2M and PTEN associated with increased overall metastatic potential. PTEN mutations were also associated with brain metastases but were enriched only in the "highly metastatic" cluster and not the brain-specific cluster. Mutations in GNAQ or GNA11 (GNA) associated with increased liver metastasis. To validate this association, we tested and demonstrated liver tropism in two GNA-mutant genetically engineered cutaneous melanoma mouse models of metastasis. Overall, our study elucidates distinct phenotypes of metastasis in patients with melanoma and identifies novel clinical and genomic associations that illuminate the drivers of clinical metastatic organotropism.

cancer biology↗

bioRxiv2022

Genomic heterogeneity and ploidy identify patients with intrinsic resistance to PD-1 blockade in metastatic melanoma

While the introduction of immune checkpoint blockade (ICB) has dramatically improved clinical outcomes for patients with advanced melanoma, a significant proportion of patients develop resistance to therapy, and mechanisms of resistance are poorly elucidated in most cases. Further, while combination ICB has higher response rates and improved progression free survival compared to single agent therapy in the front line setting, there is significantly increased toxicity with combination ICB, and biomarkers to identify patients who would disproportionately benefit from combination therapy vs aPD-1 ICB are poorly characterized. To understand resistance mechanisms to single vs combination ICB therapy, we analyze whole-exome-sequencing (WES) of pre-treatment tumor and matched normals of 4 cohorts (n=140) of previously ICB-naive aPD-1 ICB treated patients. We find that high intratumoral genomic heterogeneity and low ploidy identify patients with intrinsic resistance to aPD-1 ICB. Comparing to a melanoma cohort from a pre-targeted therapy and ICB time period ("untreated" cohort), we find that genomic heterogeneity specifically predicts response and survival in the ICB treated cohorts, but not in the untreated cohort, while ploidy is also prognostic of overall survival in the "untreated" (by targeted therapy or ICB) group. To establish clinically actionable predictions, we optimize a simple decision tree using genomic ploidy and heterogeneity to identify with high confidence (90% PPV) a subset of patients with intrinsic resistance to and significantly worse survival on aPD1 ICB treatment. We then validate this model in independent cohorts, and further show that a significant proportion of patients predicted to have intrinsic resistance to single agent aPD-1 ICB respond to combination ICB, which suggests that nominated patients may benefit disproportionately from combination ICB. We further show that the features and predictions of the model are independent of known clinical features and previously nominated molecular biomarkers. These findings highlight the clinical and biological importance of genomic heterogeneity and ploidy, and sets a concrete framework towards clinical actionability, broadly advancing precision medicine in oncology.

cancer biology↗

bioRxiv2022

CanSig: Discovering de novo shared transcriptional programs in single cancer cells

Single-cell RNA sequencing (scRNA-seq) facilitates the discovery of gene signatures that define cell states across patients, which could be used in patient stratification and drug discovery. However, the lack of standardization in computational methodologies to analyse these data impedes the reproducibility of signature detection. To address this, we developed CanSig, a comprehensive benchmarking tool that evaluates methods for identifying transcriptional signatures in cancer. CanSig integrates metrics for batch correction and biological signal conservation with a gene signature correlation metric to score according to rediscovery, cross-dataset reproducibility, and clinical relevance. We applied CanSig to ten methods and to ten scRNA-seq datasets from four human cancer types--glioblastoma, breast cancer, lung adenocarcinoma, and cutaneous squamous cell carcinoma-- representing 116 patients and 105,000 malignant cells. Our results identify BBKNN as a leading method. We showed that the signatures identified with these methods correlate with clinically relevant outcomes, including patient survival and lymph node metastasis. Thus, CanSig establishes a standardized framework for reproducible cancer transcriptomics analysis.

bioinformatics↗