Back

Earth-Observation and Environmental Vision Transformers Reveal Genome-Environment Associations in Macroalgae

Mystikou, A.; Nelson, D. R.; El Assal, D. C.; Jaiswal, A. K.; Sultana, M.; Rad-Menendez, C.; Al-Mansoori, N.; Elias, S. T.; Abdul-Hamid, M.-s.; Nayfeh, L.; Drou, N.; Burt, J. A.; Green, D. H.; Salehi-Ashtiani, K.

2025-12-30 genomics
10.64898/2025.12.30.696986 bioRxiv
Show abstract

Macroalgae thrive in extreme environments, yet the genomic basis of their tolerance remains poorly resolved. We describe nine Arabian Gulf macroalgae and integrate them with 117 published genomes (126 total; 70 Rhodophyta, 43 Ochrophyta, 13 Chlorophyta) to test genome-environment associations. Google Earth Engine (GEE) for broad-scale oceanography and 10-meter resolution AlphaEarth Foundations (AEF) embeddings for fine-scale habitat heterogeneity. We identified 157 significant (FDR q < 0.05) correlations with global GEE variables--including a strong negative temperature association with DUF3570--while AEF embeddings uncovered over 1,000 lineage-specific signals within Rhodophyta and identified climate-driven Pfam modules. The von Willebrand factor type-A domain emerged as uniquely robust across all frameworks and enriched in Arabian Gulf species. In the Arabian Gulf, enrichment of this domain is consistent with selection for adhesion mechanisms capable of withstanding chronic hydrodynamic stress compounded by high temperature and salinity. These results demonstrate that converging remote sensing with deep learning identifies conserved and lineage-specific genomic signatures of ecological differentiation across diverse macroalgal lineages.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.