Training-free protein sequence embedding in complex space using FFT.
ProtSEC is a training-free protein sequence embedding method in complex space using the Fast Fourier Transform (FFT). Unlike protein language models (ESM2, ProtBERT, ProtT5), which require GPUs for inference, ProtSEC generates embeddings on CPU with substantially lower runtime while exceeding their performance on sequence similarity search.
ProtSEC is a training-free protein sequence embedding method in complex space using Fast Fourier Transform (FFT). Unlike protein language models (ESM2, ProtBERT, ProtT5), which require GPUs for inference, ProtSEC generates embeddings on CPU with substantially lower runtime while exceeding their performance on sequence similarity search.
@article{protsec2025,title={ProtSEC: Ultrafast Protein Sequence Embedding in Complex Space Using Fast Fourier Transform},author={Raju, Rajan Saha and Islam, Rashedul},journal={bioRxiv},year={2025},month=oct,}