Fast and accurate construction of multiple sequence alignments from protein language embeddings with ARIES.

ARIES achieves higher accuracy than existing methods in constructing multiple sequence alignments, especially in low-identity regimes, with near-linear scalability across diverse protein datasets.

• Why it matters: Accurate MSAs are essential for understanding protein structure and evolution, but traditional algorithms struggle with sequences of low similarity, limiting their effectiveness in many biological analyses.
• What they did: The study developed ARIES, an algorithm that uses protein language model embeddings and a reciprocal similarity metric to align sequences via dynamic time warping, tested on multiple benchmark datasets.
• The result: ARIES outperforms current approaches in accuracy, particularly in challenging low-identity cases, and demonstrates scalable performance, indicating PLMs' potential to revolutionize comparative sequence analysis.