AI protein design starts with a desired structure or function and works toward candidate molecules; protein structure prediction starts with an amino-acid sequence and estimates the structure it may form. A typical design workflow generates a backbone, selects a sequence for it, filters candidates computationally, then tests selected designs in the lab. Each stage answers a different question, and a promising computer model is not proof that a protein will work.
Protein design and structure prediction start with different inputs
In structure prediction, the input is a protein sequence and the goal is to estimate its three-dimensional structure. AlphaFold’s 2021 paper describes predicting coordinates from an amino-acid sequence and aligned homologous sequences, and reports evaluation in CASP14, a blind assessment against newly solved structures. Those results concern structure prediction—not the success rate of proteins designed by AI. Nature’s AlphaFold paper explains the method and evaluation.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
The Color of North: The Molecular Language of Proteins and the Future of Life | $23.30 | Buy on Amazon |
| 2 |
|
Proteins: Structure and Function | $48.56 | Buy on Amazon |
| 3 |
|
Protein Chemistry (De Gruyter Textbook) | $55.36 | Buy on Amazon |
| 4 |
|
Protein: The Making of a Nutritional Superstar | $26.89 | Buy on Amazon |
| 5 |
|
Proteins: Structures and Molecular Properties | $89.98 | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
Design runs in the other direction: researchers specify a desired property, such as a particular fold, a binding interaction, a symmetric assembly, or a functional motif displayed on a stable scaffold, and use computational methods to propose candidate structures or sequences. In practice, prediction methods can be used during design to assess candidates, but prediction and design remain distinct tasks.
How a design moves from a goal to a candidate
- Define the design task. Researchers specify what the candidate should do or how it should be organized—for example, bind a target, form an assembly, or present a functional motif in a stable scaffold. The precision and type of constraints depend on the task.
- Generate a candidate backbone. RFdiffusion begins with random residue frames and iteratively denoises them toward a plausible protein backbone, conditioned on the design task. It generates a structural framework rather than directly proving that a working protein exists.
- Design a sequence for that backbone. In the RFdiffusion workflow, ProteinMPNN proposes amino-acid sequences intended to encode the generated structure. Researchers can sample multiple sequences for a design, creating candidates to assess rather than a single guaranteed answer.
- Filter candidates computationally. Structure-prediction systems can estimate whether a proposed sequence is likely to fold into a shape resembling its intended design. The RFdiffusion study used AlphaFold2-based criteria for in-silico evaluation. Passing such a filter is a reason to prioritize a candidate, not experimental confirmation.
- Make and test selected candidates. Researchers produce selected designs and characterize them experimentally to determine whether they can be made and whether they have the intended structure or function. The RFdiffusion authors report experimental characterization across designed assemblies, metal-binding proteins, and binders. The RFdiffusion paper describes its generation, sequence-design, computational evaluation, and experimental work.
What computational checks establish—and what they do not
A predicted structure that resembles the intended design helps answer whether a sequence appears compatible with that structural goal in a model. It does not establish that the protein will be expressed, remain stable, bind its intended target, or perform a biochemical function. Those are physical outcomes that require experimental characterization.
#1 Best Overall
The evidence also depends on what was actually measured. A structural result can support a structural claim; a binding assay can support a binding claim under its test conditions. Neither should automatically be treated as proof of every other desired property. There is no single score that makes a sequence generator, an inverse-folding method, a structure predictor, and a laboratory assay interchangeable.
How to read the evidence for AI-designed proteins
Published experimental successes show that particular designs can work under particular conditions; they do not establish a universal design-to-lab success rate. The RFdiffusion study reports a cryogenic electron microscopy structure of a designed binder bound to influenza haemagglutinin that the authors say is nearly identical to the design model. This is a specific structural example, not a field-wide success-rate estimate.
Rank #2
When evaluating a claim about a designed protein, distinguish what the method took as input, what it generated, and what evidence supports the result. A benchmark on a structure-prediction test set, a computational filter, and an experimental assay answer different questions. The AlphaFold CASP14 evaluation, for example, assesses prediction performance on its specified test set; it does not measure the fraction of AI-designed proteins that work in a lab.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Where AlphaFold 3 fits
AlphaFold 3 extends structure prediction to joint structures involving proteins and other molecular types, including nucleic acids, small molecules, ions, and modified residues. Its 2024 paper describes a diffusion-based architecture for predicting these biomolecular complexes. That capability can inform interaction modeling, but it is still prediction: it does not replace experiments that establish whether a proposed molecule can be made or performs the intended function. The AlphaFold 3 paper describes its approach.
Rank #3
What predicted-structure databases contain
The AlphaFold Protein Structure Database is an expanding resource of predicted structures. A database entry should be described as a prediction unless independent experimental evidence establishes that the structure was experimentally solved. The database is useful for exploring predicted structures, but its existence does not change the distinction between a model and a laboratory result. The AlphaFold Protein Structure Database provides information about its contents.
Quick Recap
Best Value
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
