Breakthroughs in computational predictions are enabling the design of novel proteins for critical medical, environmental, and technical applications. The final “designer proteins” for these applications are a product of an iterative “design, make, test, analyze” cycle until a protein with the desired properties is generated.

Currently, these campaigns are very labor intensive; typically, lasting several years and requiring contributions from many researchers. While protein purification itself is a lot of work, there are methods for enhancing the throughput on this part of the cycle.

But with recent advances, we can envision a future instrument that accelerates all of the steps in this cycle and seamlessly transitions from one step to another until a set endpoint is met.

Such an instrument holds the potential to cut down years of intense effort from research teams to mere days for a single scientist. An instrument of this type would be a real catalyst in producing the next generation of protein solutions for pressing challenges.

To be clear, this instrument is not some far-off imagination of science fiction. There are currently robust methods for achieving all of the individual steps conducted by this instrument. What’s missing is the seamless integration from one step to another, and from one cycle to another. These gaps play into AI’s strengths and represent an area where computational advances can accelerate automation to reach previously inaccessible scales of purification (da Silva, 2024).

How will proteins be purified in the future?

Now, let’s picture a future instrument that can take an input question, and iteratively cycle through the “design, make, test, analyze” cycle independently until it reaches the desired outcome.

The AI technology already exists to help design proteins for a given function (Boyoglu-Barnum et al, 2021; Cui et al, 2024; Lipsh-Sokolik et al, 2023; Silva et al, 2019; Vázquez Torres et al, 2025; Walls et al, 2020; Wolf et al, 2015; Yang et al, 2024; Zhang et al, 2025).

In our futuristic instrument, once this design stage is completed this would kick off DNA synthesis for the designed genes, which would then be automatically transformed into bacteria which are grown and induced with IPTG to initiate protein expression.

The cells would be lysed and purified using the techniques we described in the previous article, currently either tip-based or plate-based formats for highest throughput, though perhaps a new format will be invented between now and the advent of our do-it-all protein purifying instrument.

The instrument will then transfer the enzymes to another chamber where it will test their activity, as well as collect and analyze the data for this activity experiment. I’m using activity in a very broad sense here to capture any kind of feature that we are optimizing these designer proteins for. That could be monitoring the catalytic activity of these proteins for a specific chemical reaction, or binding to another protein, nucleic acid, or small molecule, for example.

Then, importantly, the instrument would automatically feed this activity information back into the design process to reiterate a new round of enzymes until a predetermined endpoint (such as 100-fold more active than the original enzyme, for example) is met (Figure 1).

Figure 1. Futuristic, fully-automated “design-make-test-cycle” that can iteratively improve protein sequences for a given function or activity until a satisfactory condition is met. 

This future instrument automatically and seamlessly connects the steps in the “design, make, test, analyze” cycle until a predetermined endpoint is met, or until it fails after a set number of cycles.

The slow steps of this cycle would be the physical steps where the DNA is being synthesized, bacteria cells are growing, the purification is happening, and the enzyme testing is taking place.

Given current computing power, the computational steps of designing enzymes and analyzing the data are actually quite quick. Though, to my knowledge, no one has yet figured out how to use AI to connect the analyze step of one run with the design step of the subsequent run.

This setup could design, make, test, and analyze around 1.5 million protein designs in a single year. While 1.5 million proteins pales in comparison to the entire protein sequence space, remember that we don’t have to get through all of that space since computation helps sort out which parts of the protein sequence universe we should be searching in.

So, that’s a vision for how AI and automation will transform protein purification to provide sufficient throughput to make advances on previously intractable problems. Since all of the individual parts of this future instrument already exist, all that is needed is a little engineering to put it all into one system and connect the individual parts together. Automated data analysis and feeding that data into the next design cycle, in particular, play into the strengths of AI and represent areas of vast time-savings compared to the current labor-intensive approaches that most labs rely on (da Silva, 2024).Such an instrument would assist basic scientific research, as well as fast-track translational research in areas such as drug discovery, materials, and environmental applications.

Related Resources

Computational Breakthroughs in Protein Prediction

High Throughput Methods for Protein Purification

References

Boyoglu-Barnum, S., Ellis, D., Gillespie, R. A., Hutchinson, G. B., Park, Y. J., Moin, S. M., Acton, O. J., Ravichandran, R., Murphy, M., Pettie, D., Matheson, N., Carter, L., Creanga, A., Watson, M. J., Kephart, S., Ataca, S., Vaile, J. R., Ueda, G., Crank, M. C., Stewart, L., … Kanekiyo, M. (2021). Quadrivalent influenza nanoparticle vaccines induce broad protection. Nature, 592(7855), 623–628. https://doi.org/10.1038/s41586-021-03365-x

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

da Silva R. G. L. (2024). The advancement of artificial intelligence in biomedical research and health innovation: challenges and opportunities in emerging economies. Globalization and health, 20(1), 44. https://doi.org/10.1186/s12992-024-01049-5

Lipsh-Sokolik, R., Khersonsky, O., Schröder, S. P., de Boer, C., Hoch, S. Y., Davies, G. J., Overkleeft, H. S., & Fleishman, S. J. (2023). Combinatorial assembly and design of enzymes. Science (New York, N.Y.), 379(6628), 195–201. https://doi.org/10.1126/science.ade9434

Norton-Baker, B., Denton, M. C. R., Murphy, N. P., Fram, B., Lim, S., Erickson, E., Gauthier, N. P., & Beckham, G. T. (2024). Enabling high-throughput enzyme discovery and engineering with a low-cost, robot-assisted pipeline. Scientific reports, 14(1), 14449. https://doi.org/10.1038/s41598-024-64938-0

Silva, D. A., Yu, S., Ulge, U. Y., Spangler, J. B., Jude, K. M., Labão-Almeida, C., Ali, L. R., Quijano-Rubio, A., Ruterbusch, M., Leung, I., Biary, T., Crowley, S. J., Marcos, E., Walkey, C. D., Weitzner, B. D., Pardo-Avila, F., Castellanos, J., Carter, L., Stewart, L., Riddell, S. R., … Baker, D. (2019). De novo design of potent and selective mimics of IL-2 and IL-15. Nature, 565(7738), 186–191. https://doi.org/10.1038/s41586-018-0830-7

Vázquez Torres, S., Benard Valle, M., Mackessy, S. P., Menzies, S. K., Casewell, N. R., Ahmadi, S., Burlet, N. J., Muratspahić, E., Sappington, I., Overath, M. D., Rivera-de-Torre, E., Ledergerber, J., Laustsen, A. H., Boddum, K., Bera, A. K., Kang, A., Brackenbrough, E., Cardoso, I. A., Crittenden, E. P., Edge, R. J., … Baker, D. (2025). De novo designed proteins neutralize lethal snake venom toxins. Nature, 639(8053), 225–231. https://doi.org/10.1038/s41586-024-08393-x

Walls, A. C., Fiala, B., Schäfer, A., Wrenn, S., Pham, M. N., Murphy, M., Tse, L. V., Shehata, L., O'Connor, M. A., Chen, C., Navarro, M. J., Miranda, M. C., Pettie, D., Ravichandran, R., Kraft, J. C., Ogohara, C., Palser, A., Chalk, S., Lee, E. C., Guerriero, K., … King, N. P. (2020). Elicitation of Potent Neutralizing Antibody Responses by Designed Protein Nanoparticle Vaccines for SARS-CoV-2. Cell, 183(5), 1367–1382.e17. https://doi.org/10.1016/j.cell.2020.10.043

Wolf, C., Siegel, J. B., Tinberg, C., Camarca, A., Gianfrani, C., Paski, S., Guan, R., Montelione, G., Baker, D., & Pultz, I. S. (2015). Engineering of Kuma030: A Gliadin Peptidase That Rapidly Degrades Immunogenic Gliadin Peptides in Gastric Conditions. Journal of the American Chemical Society, 137(40), 13106–13113. https://doi.org/10.1021/jacs.5b08325

Yang, E. C., Divine, R., Miranda, M. C., Borst, A. J., Sheffler, W., Zhang, J. Z., Decarreau, J., Saragovi, A., Abedi, M., Goldbach, N., Ahlrichs, M., Dobbins, C., Hand, A., Cheng, S., Lamb, M., Levine, P. M., Chan, S., Skotheim, R., Fallas, J., Ueda, G., … Baker, D. (2024). Computational design of non-porous pH-responsive antibody nanoparticles. Nature structural & molecular biology, 31(9), 1404–1412. https://doi.org/10.1038/s41594-024-01288-5

Zhang, Y., Srijay, D., Quinn, Z., Chatterjee, P. (2025) Metalorian: De Novo Generation of Heavy Metal-Binding Peptides with Classifier-Guided Diffusion Sampling. bioRxiv. https://doi.org/10.1101/2025.07.10.664242

• Share :
• Deel
• X
• LinkedIn

---