A new way to translate protein sequences into nucleic acid sequences
The PINATA model uses evolutionary data from related species to create more natural sequences.
Researchers introduced PINATA, which stands for Phylogenetically Informed protein to Nucleic Acid Translation with sequence Alignment. This model translates protein sequences into nucleic acid sequences. It generates sequences that mirror naturally-occurring sequences better than existing models.
The method uses multiple sequence alignments (MSAs) from related species to guide predictions. MSAs are sequences from related species. The architecture learns species-and protein specific context. It leverages evolutionary context to make biologically-informed predictions.
PINATA was evaluated across 61,934 gene trees and 882 species. It improves exact native codon recovery by 17.95 percentage points over existing models. It also preserves natural rare-codon nuances and matches native sequence properties.
The model maintains strong predictive performance even with limited evolutionary data. Previous methods treated sequences from different species in isolation. This new method leverages evolutionary relationships between nucleic acids and proteins across species.
Key facts
- PINATA improves exact native codon recovery by 17.95 percentage points.
- The model was tested using 61,934 gene trees and 882 species.
- PINATA uses multiple sequence alignments to guide its predictions.