Enzyme Design Overview

The plan is to develop an enzyme design pipeline to create enzymes that can break down harmful molecular "junk" inside cells. These accumulated wastes (such as certain intracellular aggregates and resistant metabolic byproducts) are a form of aging damage that cells cannot naturally remove. In fact, the buildup of such "garbage" inside cells is identified as one of the seven primary categories of aging damage in the SENS damage-repair approach to aging (see science).

Target Selection: The first step is choosing what molecular "junk" to target. The Targets page provides an overview of various intracellular damage targets that accumulate with age – for example, oxidized cholesterol derivatives, misfolded protein aggregates, or cross-linked waste products. Each entry in the Targets table describes one damage. The table should help prioritize neglected but important targets. Emphasis is placed on targets that (1) are significantly harmful or abundant in aging cells, (2) are not being adequately addressed by existing efforts, and (3) appear enzymatically addressable (meaning an enzyme could plausibly be designed to break them down). Focusing on such viable targets ensures that the designed enzyme makes as much impact as possible.

AI-Driven Design Pipeline: Once a target is selected, the next phase is designing an enzyme to degrade it. Thanks to recent advances in AI and computational biology, this process can largely be done in silico (on computers). The plan is to integrate many cutting-edge tools into a unified enzyme design pipeline (see the Tools page for details on existing enzyme design tools). For example, DeepMind's AlphaFold is a revolutionary AI system that accurately predicts protein structures, which helps in modeling how a candidate enzyme will fold and what it might look like. Another example is RFdiffusion, a generative neural network from the Baker Lab that can design new protein structures; traditionally, creating a working enzyme was like finding a needle in a haystack – researchers might have to test tens of thousands of engineered proteins to find one that works, but new AI methods like RFdiffusion have dramatically improved success rates (in some cases finding a viable design after only a handful of tries). By combining tools like these with others (for protein sequence design, active site modeling, molecular docking, etc.), the pipeline can generate a multitude of enzyme designs, then computationally evaluate and filter them to pick the most promising candidates before any wet-lab experiments are needed. Many of these AI tools are freely available (often open-source), and the pipeline itself is planned to be an open-source project as well – meaning it can continuously incorporate the latest techniques and even allow contributions or use by other researchers. This computational-first approach makes enzyme development far more efficient than traditional trial-and-error protein engineering, greatly narrowing down candidates in silico so that only the most promising enzymes are taken forward for experimental testing.

Tools and Pipeline Integration: The Tools page lists modern AI and computational biology tools that play roles in enzyme design. These range from structure predictors (like AlphaFold and RoseTTAFold) to generative design algorithms (like RFdiffusion and others), and from sequence optimization tools to software for simulating enzymatic reactions. Each tool in the list is described with its purpose, origin (e.g. academic lab or company), availability (open-source, free, or commercial), and how it can aid the enzyme design process. The enzyme design pipeline will leverage multiple tools in sequence or in combination – for instance, one tool might propose backbone structures for an enzyme, another might refine the enzyme's amino acid sequence for stability, and yet another might predict how well the enzyme could bind and cleave the target molecule. New tools are emerging rapidly in the field of protein engineering, so the plan is to design the pipeline to be flexible and updatable. As new AI models or algorithms become available, they can be integrated into the workflow. If certain gaps are identified (for example, a need for a specialized activity prediction that no current tool provides), the plan is to either customize existing tools or develop new software components to fill those needs. By maintaining this pipeline as an evolving, possibly open-source platform, Fix Aging AI aims to not only develop its own therapies but also provide a resource that others in the field can use and build upon.

Overcoming Challenges: Designing an enzyme that works in the human body is a complex task, and there are many challenges beyond just finding an enzyme that can break down the target molecule. The Challenges page goes into detail about these hurdles and the strategies to address them. Successfully bringing an enzyme therapy from computer design to the clinic means solving all of these issues, not just finding an active enzyme in principle. This is why the enzyme design pipeline project also involves studying and incorporating solutions for delivery, immunogenicity, regulatory considerations, and more.

Development Roadmap: After designing and refining enzyme candidates in silico, the most promising designs move into laboratory and clinical development stages. Below is an overview of the envisioned pipeline from start to finish:

Identify a Target: Select an intracellular damage molecule to focus on (from the Targets list) based on clear criteria – it accumulates with age, causes harm, and is chemically suitable for enzymatic breakdown. Defining a good target is crucial, as it sets the stage for all downstream design efforts.
In Silico Enzyme Design: Using the AI-driven pipeline, generate a large pool of candidate enzymes that could break down the target. This includes designing protein structures and sequences with potential activity against the target molecule.
Computational Filtering & Optimization: Apply additional computational analyses to evaluate and narrow the candidates. For example, use structure prediction (to check that candidates fold correctly), molecular docking or molecular dynamics (to see if the enzyme can bind the target), and other predictive models (to assess stability, specificity, and potential immunogenic sites). This step winnows down the list to a few top candidates that are most likely to succeed.
In Vitro Testing: Synthesize genes or peptides for the top candidate enzymes and test them in the lab (in test tubes and cell cultures). In vitro experiments will confirm whether the enzyme actually binds to and breaks down the target substance as intended. Data from these tests can also validate the computational predictions and guide any further refinements.
In Vivo Studies: If an enzyme proves effective in vitro, the next step is to test it in living organisms (typically in small animal models, such as mice). These preclinical trials assess how the enzyme behaves in a complex living system: does it reach the target cells, does it remain active, and are there any side effects or immune reactions? Success in animal studies provides critical evidence of safety and efficacy before proceeding to human trials.
Clinical Trials: Finally, a successful enzyme therapy would move into clinical trials in humans. This stage is the most resource-intensive, requiring rigorous testing through Phase I (safety), Phase II (efficacy in small groups), and Phase III (larger-scale efficacy and monitoring) trials. Given the high cost and scale of clinical trials, a common strategy is to partner with or license the therapy to a larger pharmaceutical company at this stage. Such partnerships can provide the funding and infrastructure needed to run human trials and bring the treatment to market. The end goal is to achieve an approved therapy that can be delivered to patients, mitigating that particular form of aging damage.