From Code to Cure: The Revolution in Drug Discovery with AI-Generated Proteins

Written by Mariana Kuznietsova | May 6, 2025 1:00:00 PM

Introduction

Artificial intelligence (AI) is rapidly becoming a driving force in therapeutic innovation. Protein design tasks are now being solved with unprecedented speed and accuracy. This overview highlights the companies, technologies, and partnerships leading this shift, and explores the opportunities and challenges shaping the future of AI-driven protein design and drug discovery.

Background/Importance

AI fosters the future of biotechnology and drug discovery, thriving on interdisciplinary collaboration of pharmaceutical giants, biotech startups, and top-tier academic institutions, which unite to expand opportunities in computational biology.

AI has opened new frontiers in protein engineering by enabling the design of tailor-made proteins. Advanced machine learning algorithms trained on vast datasets of biological sequences and structural information are used to create AI-generated proteins. SCUBA-D, a powerful AI-driven method for structure-guided de novo protein design, confirmed by the X-ray structures of proteins and a protein complex¹.

Cutting-edge tools like ProtGPT2, ProGen2, and ESM-2 generate de novo protein sequences and increasingly integrate with state-of-the-art 3D structure prediction capabilities^2,3,4. These sequence-based design approaches are usually evaluated against or complemented by structure prediction deep learning tools such as AlphaFold2 and RoseTTAFold ^5,6.

In recognition of these groundbreaking advances, David Baker, Demis Hassabis, and John Jumper were awarded the Nobel Prize in Chemistry in 2024 for their pioneering work in computational protein design and protein structure prediction, highlighting a transformative moment for biological research.

AI is no longer just a supporting tool. It’s becoming a creative partner in scientific discovery, where computational power and biological insight redefine innovation.

Startups in AI Biology

A new generation of AI-native companies is rapidly innovating in therapeutic protein design. AI protein design competitions foster innovation and accelerate research by offering immediate experimental validation, thus empowering scientists to move faster from idea to insight⁷.

EvolutionaryScale (2024): A standout startup that raised $142 million from top investors like Lux Capital, Nat Friedman, Daniel Gross, Amazon, NVentures (NVIDIA’s venture capital arm). It is behind ESM3, a next-generation language model capable of simulating 500 million years of protein evolution, with applications in generating entirely new protein functions and structures⁸.
Adaptyv Bio (2024): launched a binder design competition to design an EGFR binder. Adaptyv Bio is a company that offers screening services, provides a platform for AI-driven protein modeling, and has pioneered competitive frameworks for protein design through global contests⁹.
Rosetta Commons & Liberum Bio (2024): Collaborate on competitive design games to foster community innovation in protein modeling, by creating de novo designs and generating novel mutations to the target¹⁰.
DenovAI (2023): is a startup founded by EMBL scientists and supported by AION Labs (backed by Pfizer, AstraZeneca, AWS, and others). It focuses on de novo therapeutic antibody design and small biologics using AI models that integrate multiple data types (sequence, structure, biophysics) to speed up early drug discovery¹¹.

Notable Directions in AI-Driven Protein Design

AI is accelerating design across multiple categories:

Toxin Neutralization. Snakebites envenomation claims lives and causes severe complications. Current antivenom treatments rely on polyclonal antibodies from immunized animals, which remain costly and not always efficient. To overcome this, scientists have applied deep learning techniques to de novo design synthetic proteins that bind and neutralize these toxins¹².
CAR-T and T-cell Therapies. Targeted protein mini-binders are being explored for advanced cell therapies in oncology. The concept is a weak but constitutive activation signal in CAR-T cells even in the absence of tumor antigens, to prevent CAR-T exhaustion after continuous CAR stimulation¹³.
Drug research: An AI-powered algorithm designs new drug molecules by analyzing the 3D structure of target proteins, creating compounds that can precisely boost or block their activity¹⁴.
Antibodies and Binding Proteins: Contests and startups are focused on novel binders and functional mimics of antibodies¹⁵.

The rapid rise of AI in protein design is a part of a larger transformation across the industry. Interdisciplinary collaborations among biotech startups, pharmaceutical companies, and academic institutions drive this field.

Partnerships

Strategic partnerships are foundational to the ecosystem’s growth:

In 2023, AstraZeneca and AbbVie partnered with Absci and BigHat Biosciences AI drug creation companies, in deals over $200 million to combine AI-driven antibody design with pharma expertise to speed up antibody drug development¹⁶.
In 2025, EMBL and ELLIS. These European powerhouses signed an MoU to strengthen AI and life sciences integration. This will enable exchange programs, joint research, and training initiatives to position Europe as a leader in AI-enhanced biology¹⁷.

Venture Companies/Incubators in AI Biology

Flagship Pioneering has raised a $3.6 billion fund to support the development of an estimated 25 breakthrough companies in human health, sustainability, and artificial intelligence in 2024. This investment strategy underscores the integration of AI into the life sciences sector to accelerate innovation¹⁸.

Flagship Pioneering unveiled Abiologics, a startup dedicated to developing "supranatural" biologics termed Synteins™. These synthetic proteins are created by leveraging generative AI and high-throughput chemical synthesis, made of D-amino acids to ensure that they bind diverse therapeutically relevant targets while remaining ultrastable. This approach aims to overcome limitations of natural biologics, potentially enabling oral administration of protein-based therapies and less frequent dosing¹⁹.
Generate:Biomedicines, a Flagship-founded company, employs machine learning algorithms to design novel proteins for therapeutic applications and has a robust pipeline across various stages of development. In 2024 Novartis inked a deal potentially worth more than $1 billion with Generate:Biomedicines to develop protein therapeutics in multiple indications²⁰.

AION Labs is an innovation hub that accelerates innovation by uniting cross-industry expertise. By creating startups that use AI and cutting-edge technologies to solve key challenges in drug discovery, Anion Lab enables secure data collaboration, minimizing startup risk and speeding up the path to market. Backed by pharma giants like Pfizer, AstraZeneca, and tech leaders like Amazon Web Services (AWS), it builds companies at the intersection of computation and biology²¹. AION Labs is advancing AI-driven drug discovery through ventures like DenovAI, which designs therapeutic antibodies from scratch using biophysics and machine learning, and ProPhet, which uses generative AI to identify active small molecules for hard-to-target proteins²².

2024. CombinAble AI, a startup from AION Labs, focuses on optimizing antibody design by integrating AI and computational methods. The startup aims to reduce the time and cost associated with developing targeted antibodies, enhancing the efficiency of therapeutic development across various diseases²³.
AION Labs, in collaboration with BioMed X, has launched a global call for applications to establish a new startup for generative AI to discover and validate novel molecular target combinations. This initiative addresses novel combinations based on disease relevance, biomarker predictability, and on-target adverse reaction risks in therapeutic areas such as oncology, immune-mediated disorders, and cardiovascular-kidney-metabolic (CKM) diseases²⁴.

Opportunities and Limitations

Opportunities

AI accelerates protein design, reducing development cycles.
Advanced tools democratize participation, allowing anyone with basic resources to contribute.
Competitions and open platforms foster global collaboration and rapid knowledge sharing.
AI/ML enables the creation of tailor-made molecules for specific therapeutic needs and lower production costs.

Limitations

High pace of AI-generated protein competition risks missteps in design, which could hinder rather than help the field’s progress.
Challenges to meet many criteria tailored to each protein design⁷.

Conclusion

Fueled by breakthroughs in protein structure prediction and functional studies, and accelerated by the rapid evolution of machine learning, an era of designer proteins is coming. Drug discovery and personalizing treatments are key market drivers, while validation bottlenecks remain a restraint. Broader adoption will depend not only on technical advances but also on how regulators respond to synthetic and AI-generated molecules. BCC Research forecasts continued growth in this space, supported by trends.

in biopharma investment, cross-sector collaborations, and increasing demand for speed and scale in drug discovery.

How BCC Research Can Support Your Strategy

BCC Research delivers trusted, data-driven intelligence that helps companies make strategic decisions. We provide insights into market dynamics and offer tailored consulting, including custom research solutions to address specific business needs.

Artificial intelligence continues to transform drug discovery and protein engineering, and BCC Research remains closely attuned to this shift. Our portfolio includes in-depth market analyses that track the expanding impact and emerging opportunities of AI across diagnostics, therapeutic development, and disease management:

HLC290A. Global Markets for Artificial Intelligence in Drug Discovery

BIO261A. AI in Clinical and Molecular Diagnostics Market

BIO196B. Artificial Intelligence (AI) in Cancer

BIO255A. Artificial Intelligence (AI) in Life Sciences Market

IFT314A. AI in Healthcare Regional Analysis Market: Middle East and North Africa

References

Liu Y, Wang S, Dong J, et al. De novo protein design with a denoising diffusion network independent of pretrained structure prediction models. Nat Methods. 2024;21(11):2107-2116.

https://doi.org/10.1038/s41592-024-02437-w

Ferruz N, Schmidt S, Höcker B. ProtGPT2 is a deep unsupervised language model for protein design. Nat Commun. 2022;13(1):4348. Published 2022 Jul 27. https://doi.org/10.1038/s41467-022-32007-7
Nijkamp E, Ruffolo JA, Weinstein EN, Naik N, Madani A. ProGen2: Exploring the boundaries of protein language models. Cell Syst. 2023;14(11):968-978.e3. https://doi.org/10.1016/j.cels.2023.10.002
Bhat S, Palepu K, Hong L, et al. De Novo Design of Peptide Binders to Conformationally Diverse Targets with Contrastive Language Modeling. Preprint. bioRxiv. 2024;2023.06.26.546591. Published 2024 Jul 22.

https://doi.org/10.1101/2023.06.26.546591

Yao J, Wang X. Artificial intelligence in de novo protein design. Med Nov Technol Devices. 26 (2025) 100366

https://doi.org/10.1016/j.medntd.2025.100366

Liu Y, Wang S, Dong J, et al. De novo protein design with a denoising diffusion network independent of pretrained structure prediction models. Nat Methods. 2024;21(11):2107-2116.

https://doi.org/10.1038/s41592-024-02437-w

Callaway E. AI has dreamt up a blizzard of new proteins. Do any of them actually work?. Nature. 2024;634(8034):532-533.

https://doi.org/10.1038/d41586-024-03335-z

Inside Precision Medicine. Published June 26, 2024.

https://shorturl.at/Ic1kW

Adaptyv Bio. Published December 25, 2024.

https://www.adaptyvbio.com/blog/po104

Rosetta Commons. Published April 19, 2024.

https://shorturl.at/f7EoE

European Molecular Biology Laboratory (EMBL). Published December 14, 2023.

https://www.embl.org/news/science/using-artificial-intelligence-to-discover-therapeutic-antibodies/

Torres SV, Valle MB, Mackessy SP, et al. De novo designed proteins neutralize lethal snake venom toxins. Preprint. Res Sq. 2024;rs.3.rs-4402792. Published 2024 May

https://doi.org/10.1038/s41586-024-08393-x