AlphaFold has transformed structural biology by predicting protein shapes from amino acid sequences with unprecedented accuracy, but its impact on actual drug discovery is more nuanced than headlines suggest. While the AI tool provides structures for over 200 million proteins, medicinal chemists still rely heavily on experimental methods for the precise binding-site details needed to design effective drugs. The technology excels at answering broad architectural questions but struggles with the atomic-level accuracy required for confident chemistry decisions.

What Exactly Does AlphaFold Predict, and What Does It Miss?

DeepMind's AlphaFold2, introduced in 2021, solved a 50-year-old challenge in structural biology by predicting a protein's three-dimensional structure from its amino acid sequence. The achievement was so significant that it earned recognition in the 2024 Nobel Prize in Chemistry. The associated AlphaFold database now covers nearly the entire human proteome and almost all therapeutically relevant targets, making structural information accessible for proteins that never had experimental structures.

However, there's a critical distinction between what AlphaFold predicts and what drug designers actually need. The tool predicts a protein's most likely folded structure, typically showing a single conformational state based on patterns learned from known structures and evolutionary sequence data. A drug, though, doesn't bind to a static average structure; it binds to a specific conformation of a specific site, often reshaping that site as it binds, and in an environment full of water, ions, and other molecules that the prediction doesn't represent.

Where AlphaFold Structures Actually Help in Drug Discovery?

Despite its limitations, predicted structures have found genuine and valuable roles in drug discovery, particularly at earlier, exploratory stages where a good model is far better than no structure at all. Understanding where AlphaFold fits into the workflow is key to using it effectively.

• Targets Without Experimental Structures: For proteins that have never been crystallized, a predicted structure provides a starting point for understanding architecture, locating likely binding sites, and forming testable hypotheses, work that was simply not possible before.
• Early Target Understanding and Triage: Early in a project, a predicted structure helps assess whether a target is tractable, where druggable pockets might be, and how it relates to known protein families, informing decisions before major resources are committed.
• Accelerating Experimental Structure Determination: Predicted models have become a powerful tool for solving experimental structures faster by providing the search model that helps determine an experimental structure from diffraction data.
• Filling Gaps in Known Structures: Predicted models can complete regions missing from an experimental structure, such as flexible loops not resolved in a crystal, giving a more complete picture for analysis.

The common thread across these applications is that predicted structures excel where the alternative is nothing, and where the question is about overall architecture rather than atomic-level binding-site detail. That represents a substantial contribution to the field, though it's a different claim from saying AlphaFold has replaced experimental structure determination in drug design.

Why Predicted Structures Aren't Reliable Enough for Chemistry Decisions?

The decisive question for medicinal chemistry is whether a predicted structure is accurate enough at the binding site to guide decisions about which molecules to make. Here the evidence calls for real caution. AlphaFold reports a per-residue confidence score, and while high-confidence regions are generally reliable for overall fold, low-confidence regions, often flexible loops, are not. A binding site located in a low-confidence region should be treated with skepticism.

Binding-site detail represents the weak point of AlphaFold predictions. Even when the overall fold is accurate, the fine detail of a binding site, including exact side-chain orientations that determine how a molecule fits, is often not predicted accurately enough for confident structure-based design. This limitation means that experimental structures remain essential for lead optimization and structure-based drug design work.

How to Use AlphaFold Effectively in Your Research Workflow

• Assess Confidence Scores First: Always check the per-residue confidence score reported by AlphaFold; high-confidence regions are generally reliable for overall fold, while low-confidence regions require skepticism and experimental validation.
• Use Predictions for Architecture, Not Binding Details: Leverage predicted structures for understanding overall protein architecture, identifying potential binding pockets, and forming initial hypotheses, rather than relying on them for precise binding-site geometry needed for chemistry decisions.
• Combine Predictions with Experimental Methods: Use AlphaFold predictions as a starting point to accelerate experimental structure determination through molecular replacement in crystallography, rather than as a replacement for experimental validation.
• Prioritize Targets Without Existing Structures: Focus on applying AlphaFold predictions to proteins that have never been crystallized, where the tool provides genuine value by offering a structure where none existed before.

AlphaFold 3, the newer version of the tool, extends prediction to protein-ligand complexes and outperforms traditional docking on static cases. However, it still struggles with large conformational changes and shows conformational biases that limit its reliability for some targets. This means that while the technology continues to improve, the fundamental gap between prediction and the atomic-level accuracy needed for confident drug design decisions persists.

The honest assessment is that AlphaFold represents the most significant structural biology tool in a generation, but it has changed drug discovery in a more targeted way than the initial excitement suggested. It has democratized access to protein structures, accelerated certain workflows, and provided starting points where none existed. For the day-to-day work of finding and designing drugs, though, experimental structural biology remains essential, particularly for the binding-site details that ultimately determine whether a drug candidate will work.