Magnus Rattray

NexuST is a hierarchical foundation model for spatial transcriptomics that improves cell-type annotation and spatial prediction by integrating gene-level and cell-level modeling.

DeepPathway predicts pathway expression from H&E images using contrastive learning, validated on prostate cancer and brain tumor datasets.

This study performs ancestry-specific TWAS across six tissues to refine T2D GWAS loci, identifying 289-322 independent signals and highlighting tissue and ancestry-specific effects.

The paper uses imaging mass cytometry to analyze ECM composition in mouse models of allergic airway inflammation, identifying distinct matrix environments associated with different cell types and tissue regions.

The paper uses explainable AI on a genomic foundation model to identify RBP motifs influencing translation and mRNA decay in Drosophila.

This study presents a transcriptomic atlas of the developing human outflow tract and its adult aortic valve derivatives, revealing persistent embryonic signatures and key regulatory factors in valve development.

PISTACHIO improves spatial cell-type deconvolution in tumours using proteomics constraints and non-negative matrix factorization.

SIMOFF infers metabolic objectives from flux data, improving gene essentiality predictions in yeast and CHO cells.

DeepSynBa predicts complete dose-response matrices for drug combinations, outperforming existing methods and enabling more precise and actionable drug synergy analysis.

CellTypeAI improves scRNA-seq cell type annotation by using local generative-AI models, outperforming conventional and cloud-based methods in accuracy and sensitivity.

This paper introduces UVAE to integrate unpaired clinical flow cytometry data, improving cell type signal and clustering by correcting batch effects.

This paper shows that using one-hot encodings for drugs and cell lines in synergy prediction models can perform as well as or better than using molecular features, suggesting models rely more on identifiers than features.

metaAPA integrates polyA site predictions from multiple tools to provide high-confidence and subset-supported sites for single-cell and spatial transcriptomics data.

This study identifies gene regulatory networks driving periderm formation in early skin development using single-cell multiome and spatial transcriptomics data.

Mantpy is a framework for analyzing the extracellular matrix in spatial proteomics by representing it as spatial graphs and integrating it with cell data for joint analysis.

CellPie is a fast, unsupervised method for factor discovery in spatial transcriptomics using joint non-negative matrix factorization, showing improved performance on human cancer datasets.

The paper uses the SunTag system to study mRNA translation regulation in Drosophila, showing leaky repression of maternal hb mRNA and demonstrating its use as a reporter for ribosome pausing and ER-localized translation.

The paper shows spatially regulated mRNA decay sharpens the even-skipped expression pattern in Drosophila embryos by improving stripe definition and affecting segmentation.

The paper evaluates APA identification tools in single-cell and spatial transcriptomics, finding no single method is optimal, and recommends tool selection based on data characteristics.

The paper introduces GPMelt, a hierarchical Gaussian process model for analyzing TPP-TR datasets, enabling unbiased analysis of unconventional melting curves and improving the detection of protein interactions.

The paper maps the spatial transcriptional changes in human liver cirrhosis, identifying gene signatures and cell state transitions linked to disease progression.

The paper investigates how transcription burst amplitude modulation mediates dosage compensation in Drosophila embryos by analyzing gene expression dynamics.

The paper estimates mRNA half-lives in Drosophila embryos using RNA-seq time series and Gaussian process regression, showing correlations between mRNA stability and P-body association.

The paper introduces a scalable method to infer transcriptional kinetic parameters from MS2-McP time series data, enabling efficient analysis of promoter activity in Drosophila embryos.

This paper introduces a Bayesian method to infer kinetic parameters of oscillatory gene regulation from single-cell time series data and demonstrates its effectiveness on murine neural progenitor cells.

The paper introduces GPcounts, a non-parametric method for modeling temporal and spatial counts data in RNA-seq using a negative binomial likelihood, showing improved performance over Gaussian and Poisson models.

The paper identifies gene targets at risk loci for skin-related disorders by integrating HiChIP data from relevant cell lines and public datasets.

The paper investigates how HOX paralogs interact with TALE TFs to generate tissue-specific enhancer activation patterns in mouse embryos.

The paper shows that BMP signaling modulates transcriptional burst frequency by altering promoter activation rate in Drosophila embryos, leading to graded mRNA output across expression domains.