Summary
Ancestral sequence reconstruction outperforms generative ML methods on functional enzyme design (1).
Details
This was evaluated using either MSA Transformer or a generative adversarial network trained on natural sequences:
Of the CuSOD sequences, none of the test or ESM-MSA and only two of the GAN sequences displayed activity. Of the MDH sequences, none of the GAN or ESM-MSA sequences had activity, but 6 of the test sequences did. In contrast, the ASR model generated 9 of 18 and 10 of 18 active enzymes for CuSOD and MDH, respectively.”
1.
Johnson SR, Fu X, Viknander S, Goldin C, Monaco S, Zelezniak A, et al. Computational scoring and experimental evaluation of enzymes generated by neural networks. Nature Biotechnology. 2024;43(3):396–405. Available from: https://doi.org/10.1038/s41587-024-02214-2