Michael A. Skinnider

This study shows that caloric restriction before tumor induction delays cancer cachexia by reducing systemic inflammation and preserving lean mass.

The paper exposes a flaw in evaluating small molecule MS/MS models by showing trivial models can perform well due to dataset biases and proposes a new benchmark to assess true model capability.

RetroMol converts natural product structures and biosynthetic gene clusters into shared primary sequences to enable biosynthetic similarity comparisons.

The paper uses in vivo isotope tracing and AI to identify biosynthetic origins of unknown metabolites, revealing new metabolic pathways and age-related changes.

This paper introduces DeepMet, a chemical language model that predicts and discovers unknown metabolites using mass spectrometry data.

The paper introduces Vespucci, a machine-learning method for identifying perturbation-responsive regions and genes in spatial transcriptomics data, showing it outperforms existing methods.

This study identifies neuronal architectures responsible for autonomic dysreflexia after spinal cord injury and demonstrates how epidural stimulation can counteract it in mice, rats, and humans.

The paper introduces the Tabulae Paralytica, a multimodal atlas of spinal cord injury, revealing biological principles and developing a gene therapy to restore walking in old mice.

This study evaluates differential expression methods in single-cell transcriptomics, finding that many produce false discoveries due to ignoring biological variation.

The paper identifies protein biomarkers for spinal cord injury severity and recovery using proteomic analysis of human and porcine samples, showing high predictive accuracy.

This study evaluates the robustness of graph-based clustering algorithms to network noise in protein-protein interaction networks and introduces a tool to assess cluster reproducibility.

The paper introduces a machine-learning method, Augur, to identify cell types most responsive to biological perturbations in single-cell data and validates it on multiple datasets.

This study maps the dynamic changes in the human protein-protein interaction network induced by interferon signaling, revealing novel regulatory mechanisms and functional contexts for interferon-stimulated genes.

The paper shows that many literature-curated protein interactions are not valid under specific experimental conditions and recommends using condition-specific subsets as gold standards.

This study maps protein-protein interactions across seven mouse tissues, revealing over 27,000 unique interactions and showing rewiring of interactions that is not predicted by gene expression.

PrInCE is a bioinformatics pipeline that predicts protein interactomes from co-elution data, offering faster and more accurate results than existing tools.