IMMUNOLOGY2026™ Conference Recordings For Attendees-Characterizing spatial functional microniches with SpaceTravLR

Video Transcription

Video Summary

Jishnu from the University of Pittsburgh presented Space Traveler, an interpretable machine-learning framework for studying spatial immunology. Unlike many prediction-focused models, Space Traveler aims to infer plausible biological mechanisms by combining transcriptional regulation, ligand-receptor signaling, and spatial context. The method uses a sparse group LASSO model with spatial weights learned by a CNN or vision transformer, enabling in silico perturbations and mechanistic predictions without training on perturbation data. Across multiple tissues and platforms, it recapitulated known immune circuits in tonsils, tumors, and lung asthma models, including TFH, B-cell, macrophage, CD8 T cell, and TRM biology. It also identified novel candidate interactions, such as CCR4 in asthma, which matched true experimental spatial perturbation results. Overall, Space Traveler appears to uncover functional spatial microniches and predict perturbation effects with strong agreement to known biology and experimental data.

Meta Tag

Date April 18, 2026 2:00 PM - 2:15 PM

Room 102

Session Computational Tool Development

Speaker Jishnu Das

Track Computational and Systems Immunology (COMP)

Year 2026

Keywords

spatial immunology

machine learning

transcriptional regulation

ligand-receptor signaling

spatial context

April 18, 2026 2:00 PM - 2:15 PM

102

Computational Tool Development

Jishnu Das

Computational and Systems Immunology (COMP)

2026