Engineering enzymes for plastic degradation, machine learning miracle or evolutionary naivety?

We and others have had great success in improving the functionality of plastic-degrading enzymes using machine learning algorithms. Enzymes such as FAST-PETase and TurboPETase can completely depolymerize a wide variety of post-consumer plastics, with activity and thermostability properties which far exceed their wildtype scaffolds (Cui et al., 2024; Lu et al., 2022). These gains in enzyme performance have enabled significant improvements in PET hydrolysis and produced enzymes that are lending credence to the possibility of economical enzyme-based recycling and upcycling (“Enzymatic recycling,” 2025). Given the often-poor hit rates that arise from machine learning approaches to protein engineering, why have plastic-degrading enzymes been so amenable to this approach? 

It likely stems in part from the evolutionary naivete of the proteins themselves.  Plastic-degrading enzymes have only recently been discovered in part because plastics themselves are a newcomer on the evolutionary timeline, with the most well degraded plastics, PET and nylon, only having been developed in the 1930s and 1940s (Gomollón-Bel, 2016). Thus, the improvements in the performance of ML-engineered enzymes may be due to the fact that evolution has only had 70-80 years to optimize plastic degradation, in comparison to the millions to billions of years of optimization that has occurred with other hydrolases, such as proteinases and cutinases (from which many of the plastic-degrading enzymes seem to be derived). Since most protein engineering machine learning approaches are trained on the vast set of proteins that have undergone millions to billions of years of evolutionary optimization, it is perhaps not surprising that this data can be used to ‘accelerate’ the evolution of plastic-degrading enzymes, albeit in silico rather than in nature. The ML algorithms are essentially engineering within the bounds of what the databases provide:  the wild-type sequences, structures, and functions of proteins (Yang et al., 2024).  But in the case of plastic-degrading enzymes, there is no wild-type per se, because the functions are still just beginning to evolve.  Thus, machine learning platforms can ‘accelerate’ evolution of plastic-degrading enzymes by relying on what they know of the evolution of all other enzymes.

Related to this insight is another evolutionary lemma, which is that evolution often moves from specialist à generalist à specialist (Kosonocky & Ellington, 2023). The ‘tween plastic-degrading enzymes are only just separating themselves from their parents, with PETases being a more generalist version of cutinases, and nylon hydrolases being (even more) generalist versions of N-terminal nucleophile hydrolases, whose substrate specificity is also quite broad.  Now that the enzymes are growing up to specialize on the new xenobiotics they’re encountering, there are many available mutations that can help them out, while likely reducing the range of substrates they can cleave.

At the root, plastics are xenobiotics, human-made compounds, and thus it should not be surprising that despite the otherwise advanced timelines of bacterial evolution the genesis of new enzymes is only beginning. Irrespective of evolutionary rationales, machine learning should prove especially useful for creating and improving enzymes that can transform xenobiotic compounds, and we anticipate that the burst of research activity in engineering these enzymes will continue to accelerate, especially as tools enabling increased experimentation and participation become democratized. Beyond industrial interest and funding, contests such as the Align Foundation’s PETase engineering should further enable the rapid development of enzymes for a whole host of problems, with the success of the project possibly marking a new model for community driven enzyme development.  

Cui, Y., Chen, Y., Sun, J., Zhu, T., Pang, H., Li, C., Geng, W.-C., & Wu, B. (2024). Computational redesign of a hydrolase for nearly complete PET depolymerization at industrially relevant high-solids loading. Nature Communications, 15(1), 1417. https://doi.org/10.1038/s41467-024-45662-9

Enzymatic recycling. (2025). Carbios. https://www.carbios.com/en/enzymatic-recycling/

Gomollón-Bel, F. (n.d.). Polyethylene terephthalate. Chemistry World. Retrieved August 12, 2025, from https://www.chemistryworld.com/podcasts/polyethylene-terephthalate/1017555.article

Kosonocky, C. W., & Ellington, A. D. (2023). Evolving to Evolve, Dan Tawfik’s Insights into Protein Engineering. Biochemistry, 62(2), 145–147. https://doi.org/10.1021/acs.biochem.2c00668

Lu, H., Diaz, D. J., Czarnecki, N. J., Zhu, C., Kim, W., Shroff, R., Acosta, D. J., Alexander, B. R., Cole, H. O., Zhang, Y., Lynd, N. A., Ellington, A. D., & Alper, H. S. (2022). Machine learning-aided engineering of hydrolases for PET depolymerization. Nature 2022 604:7907, 604(7907), 662–667. https://doi.org/10.1038/s41586-022-04599-z

Yang, J., Li, F.-Z., & Arnold, F. H. (2024). Opportunities and Challenges for Machine Learning-Assisted Enzyme Engineering. ACS Central Science. https://doi.org/10.1021/acscentsci.3c01275

How good is ML protein engineering? 

One of the conundrums facing protein engineers is whether (and how) to use directed evolution versus machine learning.  On the one hand, directed evolution will generally get you where you’re going, as most selections and screens are directly for function.  However, the path can be circuitous and inefficient (as above).  In contrast, machine learning holds the promise of providing a more direct path to function, predicting the mutations that should work, rather than randomly chancing upon them.  However, there are a number of hurdles that must be traversed (also ala above), and some of these hurdles are far from easily cleared. 

Still, the excitement attending the use of machine learning for protein engineering is well-deserved.  We continue to utilize the package MutCompute (mutcompute.com), which is quite old (in the general scheme of things), but still proves quite adept at predicting mutations that generally improve protein function (most recently, photocatalysts (Liu et al., 2024). However, the hit rate for mutations for many software packages can be relatively low, requiring considerable experimental efforts to identify a few variants that improve function. For instance, MutCompute was used to select 159 predicted beneficial variants of the plastic-degrading enzyme PETase, and ultimately only four were used for further combinations to make a much better enzyme (Lu et al., 2022). Researchers used three protein generative models to redesign malate dehydrogenase and copper superoxide dismutase (Johnson et al., 2025). Trained on 5000 protein sequences per enzyme, the models generated 144 variants for experimental testing, with only 19% showing activity above background. However, additional computational filtering improved success rates by 50–150%.

But not all proteins are equal in terms of predictions, and in particular, it has proven exceedingly difficult to generate predictions for allosteric proteins. Rosetta was used to engineer LacI to respond to new ligands using computational design of the ligand binding pocket. Results showed that 14 out of 15 of the highest-ranked designs did not repress transcription; therefore, they focused on screening variants (in the range of thousands) with fewer mutations. They did end up identifying variants that responded to each of the new ligands, but only because a high-throughput selection-screening method was used to evaluate function (Taylor et al., 2016).

In our recent bioRxiv study (Clark-ElSayed et al., 2025), we tested whether new generative protein design models could be used to engineer allosteric transcription factors. Specifically, we compared LigandMPNN–a structure-informed generative –with traditional directed evolution for engineering RamR to respond to benzylisoquinoline alkaloids. Using two structure-prediction tools, we generated protein-ligand complexes and used these as backbones to LigandMPNN, targeting the same residues as in the directed evolution screen. We cloned and tested nine designed variants in E. coli, but none were able to repress transcription and did not respond to the target molecules. This demonstrates the current limitations of computational design for changing allosteric binding specificity.

So, hit rates may be especially low for allosteric proteins in part because most algorithms do not explicitly take into account conformational change. While one might expect that sequence-based methods like ESM should implicitly account for such constraints, they have also struggled to engineer allosteric transcription factors, potentially because they lack awareness of long-range residue interactions. To address this, protein design models could be integrated with tools that identify or preserve functionality. For example, in a recent study, FuncLib was used to design a library of approximately 17,000 TtgR variants, of which ~85% retained transcriptional repression activity. FuncLib combines evolutionary and energy-guided design to introduce mutations that increase thermodynamic stability, enabling it to explore sequence space while preserving allosteric function, thereby enhancing the probability of generating functional designs (Nishikawa & Chen, 2024).

References

Clark-ElSayed, A., Creed, E., Nayvelt, K., & Ellington, A. (2025). Comparing LigandMPNN and Directed Evolution for Altering the Effector-Binding Site in the RamR Transcription Factor. Synthetic Biology. https://doi.org/10.1101/2025.07.10.663684

Johnson, S. R., Fu, X., Viknander, S., Goldin, C., Monaco, S., Zelezniak, A., & Yang, K. K. (2025). Computational scoring and experimental evaluation of enzymes generated by neural networks. Nature Biotechnology, 43(3), 396–405. https://doi.org/10.1038/s41587-024-02214-2

Liu, Y., Bender, S. G., Sorigue, D., Diaz, D. J., Ellington, A. D., Mann, G., Allmendinger, S., & Hyster, T. K. (2024). Asymmetric Synthesis of α-Chloroamides via Photoenzymatic Hydroalkylation of Olefins. Journal of the American Chemical Society, 146(11), 7191–7197. https://doi.org/10.1021/jacs.4c00927

Lu, H., Diaz, D. J., Czarnecki, N. J., Zhu, C., Kim, W., Shroff, R., Acosta, D. J., Alexander, B. R., Cole, H. O., Zhang, Y., Lynd, N. A., Ellington, A. D., & Alper, H. S. (2022). Machine learning-aided engineering of hydrolases for PET depolymerization. Nature, 604(7907), 662–667. https://doi.org/10.1038/s41586-022-04599-z

Nishikawa, K., & Chen, J. (2024). Highly multiplexed design of an allosteric transcription factor to sense novel ligands [Dataset]. Zenodo. https://doi.org/10.5281/ZENODO.13381000

Taylor, N. D., Garruss, A. S., Moretti, R., Chan, S., Arbing, M. A., Cascio, D., Rogers, J. K., Isaacs, F. J., Kosuri, S., Baker, D., Fields, S., Church, G. M., & Raman, S. (2016). Engineering an allosteric transcription factor to respond to new ligands. Nature Methods, 13(2), 177–183. https://doi.org/10.1038/nmeth.3696