New AI Tool Could Replace Costly Cancer Gene Expression Profiling

Why It Matters

By slashing time and cost, Path2Space could democratize spatial genomics, accelerating personalized oncology and biomarker discovery at scale.

Key Takeaways

• Path2Space predicts spatial expression of ~5,000 genes from biopsy images.
• Predictions match measured expression across three independent breast cancer cohorts.
• Processing takes minutes, cutting costs from thousands to under $100 per sample.
• Allows analysis of thousands of slides, accelerating spatial biomarker discovery.
• Adaptable to other tumor types; head‑and‑neck trial launching soon.

Pulse Analysis

Spatial transcriptomics has emerged as a powerful lens for understanding tumor heterogeneity, but its adoption has been hampered by expensive reagents and lengthy workflows. Traditional platforms require specialized equipment and can cost several thousand dollars per sample, limiting studies to small cohorts. Path2Space flips this paradigm by leveraging high‑resolution digital pathology images—already a routine part of cancer diagnostics—to predict gene‑expression patterns using deep learning. This approach not only compresses a multi‑week process into minutes but also brings the per‑sample price down to under $100, making large‑scale spatial analyses financially feasible for academic labs and community hospitals.

The model’s performance was rigorously benchmarked on three separate breast‑cancer datasets, where predicted expression of nearly 5,000 genes correlated closely with actual spatial sequencing results. Such accuracy enables researchers to map spatial biomarkers that influence treatment response, a task previously constrained by the scarcity of affordable data. By unlocking thousands of archived pathology slides for spatial analysis, Path2Space opens a new frontier for retrospective studies, potentially revealing novel therapeutic targets and risk stratification tools without the need for fresh tissue collection.

Looking ahead, the technology’s flexibility suggests rapid adaptation to other tumor types, with a head‑and‑neck cancer trial already in preparation. If clinical trials confirm its predictive utility, Path2Space could become a standard companion diagnostic, guiding precision therapies and informing trial enrollment. The cost‑effective, AI‑driven workflow also positions it as an attractive asset for biotech firms seeking to integrate spatial genomics into drug development pipelines, potentially reshaping the economics of personalized cancer care.