Protein Structure Prediction

Published

August 22, 2026

Protein structure prediction with AI estimates three-dimensional coordinates from amino-acid sequence. AlphaFold 2 made single-chain prediction a routine research input at CASP14, with median backbone accuracy comparable to experimental structures for many well-folded domains (Jumper et al., 2021). The AlphaFold Protein Structure Database now covers over 214 million predicted structures (Varadi et al., 2024). AlphaFold 3 extends the task to complexes with nucleic acids, ions, and small molecules (Abramson et al., 2024); Boltz-1 and Chai-1 are open preprint alternatives for that broader scope (Wohlwend et al., 2024, preprint; Chai Discovery et al., 2024, preprint). The hard questions are which molecular state a prediction represents, how to read pLDDT, PAE, and ipTM, and what an experiment still has to prove.

Learning Objectives

Use this chapter to:

  • Infer useful three-dimensional protein and biomolecular structures when experimental structures are unavailable, incomplete, or too slow for the research question.
  • Judge whether a predicted structure is usable from its confidence scores, conformational state, disorder, ligands, nucleic acids, complexes, and assay context.

Prerequisites: AI for the Life Sciences recommended, Foundation Models for Biology helpful for the ESMFold and ESM-3 sections.

Summary: Infer useful three-dimensional protein and biomolecular structures when experimental structures are unavailable, incomplete, or too slow for the research question. Single-chain structure prediction is routine for many proteins; complexes, ligand interactions, dynamics, and restricted releases remain more complicated.

Key point: Confidence scores, conformational state, disorder, ligands, nucleic acids, complexes, and assay context decide whether a predicted structure is usable. Open question: how well predictions handle dynamics, ligands, complexes, and decisions that depend on more than a static fold.

Bottom line: Structure prediction feeds protein design, antibodies, variant interpretation, vaccine antigen work, chemical biology, and small-molecule discovery.

Field Guide

What is this field trying to solve? Estimate three-dimensional protein and biomolecular structure when experimental structures are unavailable, incomplete, or too slow for the research question.

What is the core idea? Read predicted coordinates together with confidence metrics and the biological state the research decision requires.

What is the current state of the field? AlphaFold 2 established high accuracy for many stable, well-folded single-chain domains. AlphaFold 3 and open systems such as Boltz and Chai extend the task to complexes that can include proteins, nucleic acids, ions, modified residues, and small molecules (Jumper et al., 2021; Abramson et al., 2024).

What do we know, and what remains open? pLDDT describes local confidence, PAE describes confidence in relative placement, and ipTM describes predicted interface quality. None of these metrics establishes function, mechanism, binding affinity, or the biological state present in a cell.

Why does this matter? Structure prediction is a strong source of hypotheses, but consequential decisions about conformational state, ligand geometry, dynamics, mechanism, or regulatory-grade characterization still require experimental evidence.

Introduction: From Grand Challenge to Routine Input

December 2020, CASP14 (online): AlphaFold 2 achieved a median backbone accuracy of 0.96 Å r.m.s.d.95 across 87 CASP14 domains, compared with 2.8 Å for the next-best method in the paper’s analysis (Jumper et al., 2021). The blind benchmark established a major improvement in single-chain structure prediction; it did not establish dynamics, disorder, complexes, ligand interactions, or biological-state selection.

July 2022: DeepMind and EMBL-EBI release the AlphaFold Protein Structure Database with predictions for approximately every cataloged UniProt protein. By the 2024 database update, coverage exceeds 214 million sequences (Varadi et al., 2024).

May 2024: AlphaFold 3 extends prediction to nucleic acids, ions, small-molecule ligands, and modified residues using a diffusion-based generative architecture (Abramson et al., 2024). Initial release is a web server with limits and without training code.

October 2024: The Royal Swedish Academy of Sciences awards the 2024 Nobel Prize in Chemistry jointly to Demis Hassabis and John Jumper (AlphaFold) and David Baker (computational protein design and Rosetta).

November 2024: MIT’s Barzilay and Jaakkola groups release Boltz-1, an open-source AlphaFold 3-class biomolecular interaction model (Wohlwend et al., 2024, preprint). The earlier open Chai-1 (Chai Discovery et al., 2024, preprint) had already established the community pattern.

These are not just modeling milestones. They are a change in the default operating assumption of structural biology: for many proteins, a usable structural model is available before any experiment is run.

The first proteome-scale use case made the practical boundary clear. AlphaFold was applied to 98.5% of the human proteome, providing high-confidence models for many well-folded domains while also marking extensive low-confidence regions that often correspond to disorder or unresolved structural context (Tunyasuvunakool et al., 2021). The value is not only the model coordinates; it is the confidence map that tells the researcher what not to trust.

The chapter that follows is organized around four questions a researcher actually faces:

  1. Which system should I use, and why?
  2. What does the confidence metric actually mean for my decision?
  3. What does the prediction not tell me?
  4. What experiment do I still owe before I commit a program decision?

The Architecture Lineage

Single-Chain Prediction (AlphaFold 2 and Successors)

AlphaFold 2 (Jumper et al., 2021) combined three ideas:

  • Multiple sequence alignment (MSA) input. Evolutionary covariation across homologs encodes structural constraints. AlphaFold 2 reads the MSA as a feature, not as input text.
  • Triangle attention and multiplicative updates. Geometric consistency is learned over pairwise residue representations through repeated Evoformer blocks.
  • End-to-end differentiable structure module. A direct geometric prediction of backbone frames and side-chain rotamers, trained with structural and auxiliary losses.

The system was open-sourced shortly after publication. RoseTTAFold (Baek et al., 2021) shipped from the Baker lab in parallel with a different “three-track” architecture and was equally consequential because it established that AlphaFold 2’s accuracy was not architecture-locked.

MSA-Free Prediction (ESMFold and OmegaFold)

For orphan proteins (no homologs), de novo designs (no evolutionary signal), or speed-critical workflows, MSA generation is the bottleneck. Two responses emerged:

  • ESMFold (Lin et al., 2023). Meta’s ESM-2, a 15-billion-parameter protein language model, predicts structure from sequence alone. The paper reported substantially faster inference than MSA-based workflows, with lower average structure accuracy than AlphaFold 2. Speed depends on sequence length, hardware, batching, and whether MSA-search time is included.
  • OmegaFold (Wu et al., 2022, preprint). Single-sequence prediction with comparable goals, useful particularly for designed proteins.

The practical rule: compare an MSA-based system with a single-sequence system when evolutionary depth is limited or turnaround time matters. Model choice depends on the target, available homologs, compute, accuracy requirement, and validation plan.

Complex and Interaction Prediction

  • AlphaFold-Multimer (Evans et al., 2021) extended AlphaFold 2 to multi-chain protein complexes, introducing the ipTM (interface predicted TM-score) metric for interface confidence.
  • AlphaFold 3 (Abramson et al., 2024) replaced the deterministic structure module with a diffusion-based generative head and extended the chemical scope to nucleic acids, ions, modified residues, and arbitrary small-molecule ligands.
  • Boltz-1 (Wohlwend et al., 2024, preprint) and Chai-1 (Chai Discovery et al., 2024, preprint) reproduced AlphaFold 3-class capabilities with permissive open-source licenses.
  • Boltz-2 (Passaro et al., 2025, preprint) added binding-affinity prediction alongside structure for the same chemistry.

Complex prediction has independent support, but the claim is narrower than “AlphaFold solves interactions.” Bryant and colleagues showed that AlphaFold2-style modeling can improve protein-protein interaction prediction when paired with appropriate ranking and confidence filters (Bryant et al., 2022). Interface confidence, stoichiometry, cellular localization, and expression context still need separate biological checks.

Why “AlphaFold 3-class” Matters Operationally

DeepMind released AlphaFold 3 inference code and model parameters for academic, noncommercial use in November 2024. That release does not provide unrestricted commercial rights or training code. Boltz and Chai use different code, weights, evidence, and licenses. Access, license, benchmark scope, and local validation must be evaluated separately.

Reading Confidence Metrics

A confident-looking AlphaFold cartoon is not evidence. The confidence outputs are.

pLDDT (predicted Local Distance Difference Test)

A per-residue score, 0 to 100, predicting how well the local atomic environment matches what an experimental structure would show. The DeepMind-recommended bands are:

Band Interpretation Practical Use
> 90 Very high local confidence Supports local-geometry hypotheses; it does not validate side-chain rotamers, pockets, or docking poses
70–90 Confident local backbone Treat side-chain and ligand-interaction claims cautiously
50–70 Low local confidence Fold and local-detail claims require orthogonal evidence
< 50 Very low local confidence May reflect disorder, flexibility, or model uncertainty; do not equate it with a measured disordered state

Disordered regions are a feature of biology, not a model failure. Low pLDDT is associated with disorder in many contexts, but it can also reflect uncertainty for other reasons. A low confidence score is not itself an experimental disorder measurement (Akdel et al., 2022).

PAE (Predicted Aligned Error)

A matrix giving expected error in residue j when the structure is aligned on residue i. The PAE matrix encodes domain structure and inter-domain confidence:

  • Block-diagonal low-PAE regions: Confident domains
  • Off-block-diagonal low-PAE regions: Confident relative orientation between domains
  • Off-block-diagonal high-PAE regions: Domain orientations are essentially unconstrained: the prediction is not a single global geometry, it is one plausible arrangement

The most common AlphaFold misinterpretation is treating a high-pLDDT multi-domain prediction as a confident full-length structure when the PAE matrix says inter-domain orientation is undetermined.

ipTM (Interface predicted TM-score): Complexes Only

For multi-chain predictions, EMBL-EBI provides these general interpretation bands (EMBL-EBI AlphaFold training):

  • ipTM > 0.8: Confident, high-quality interface prediction
  • ipTM 0.6–0.8: Uncertain; inspect chain-pair confidence, PAE, disorder, and alternative predictions
  • ipTM < 0.6: The complex may be incorrect; do not commit consequential experimental design without orthogonal evidence

These are model-confidence guides, not interaction or affinity thresholds. Combine ipTM with pTM, PAE, pLDDT, chain-pair metrics, biological context, and experimental evidence.

What Structure Prediction Does Not Solve

Capability Status Why
Single-chain stable fold Largely solved for well-folded domains AlphaFold 2 + AFDB cover the majority of the structural proteome
Conformational ensembles Open problem A prediction is one state; many proteins occupy multiple functionally distinct conformations
Apo vs. holo geometry Partially Predictions tend to bias toward training distribution; cryptic pockets that only open with ligand are systematically missed
Allosteric and signaling states Open Activation states (GPCRs, kinases, ion channels) require additional context: sometimes co-folding with binding partners
Intrinsically disordered regions Predictable as disordered, not as structures The model honestly reports low pLDDT; reading this as “wrong” is the user’s error
Ligand chemistry beyond training Limited AF3-class systems extend chemical scope but novel ligand classes, covalent binders, and metal coordination remain hard
Membrane environments Partial Predictions can be plausible without an explicit membrane; lipid interactions and oligomeric state in-membrane often need experimental support
Hydrogen positions and water networks No Out of scope for these systems
Free-energy landscapes and kinetics No Static structures are not energetics; coupling with MD remains necessary
Mechanism No A structure is geometry, not function. Mechanism needs perturbation evidence.
A Specific Trap: “AlphaFold-Validated”

A vendor claim that a molecule was “AlphaFold-validated” against a target is, on its face, meaningless. AlphaFold predicts structure; it does not validate molecules. The honest version of the claim is: “we docked our molecule against an AlphaFold-predicted structure of the target and observed favorable scoring.” That claim, in turn, is one input to a validation campaign: not a substitute for binding, cell-based, and pharmacological evidence.

Decision Framework: Choosing a System

Situation First Choice Why
Single-chain protein with deep MSA, decision-relevant AlphaFold 2 (or RoseTTAFold for open code) Highest accuracy on the canonical task
Orphan / designed / fast-iteration single-chain ESMFold (or OmegaFold) MSA-free, fast
Protein-protein complex, moderate confidence acceptable AlphaFold-Multimer Established, well-characterized failure modes
Protein + ligand / NA / ion, cloud-acceptable AlphaFold 3 web server Highest-published accuracy on the AF3 task set
Protein + ligand / NA / ion, on-premise required Compare Boltz, Chai, and other locally deployable options Verify target-specific evidence, current licence, artifacts, and compute requirements
Binding-affinity ranking matters Boltz-2 (with caution and orthogonal validation) Adds affinity head; still vendor-class evidence
Antibody-antigen interface AlphaFold 3 / Boltz / Chai with explicit caveats Antibody loops remain a known weakness for all systems
Membrane protein in lipid environment Any of the above + MD + experimental data No system claims to handle membrane partitioning natively

The deeper rule: choose the system by the biology and the access constraints, not by the brand of the model.

The AlphaFold 3 Open-Access Episode

AlphaFold 3 was published in Nature in May 2024 (Abramson et al., 2024). The initial release used a hosted server with usage and commercial restrictions and did not include code or weights. DeepMind released inference code and model parameters for academic, noncommercial use in November 2024, and Nature published an addendum describing the change (Abramson et al., 2024). Training code and unrestricted commercial rights were not part of that release.

The response was infrastructural, not rhetorical:

In October 2024, the Nobel Prize in Chemistry was awarded jointly to Demis Hassabis and John Jumper for protein structure prediction and to David Baker for computational protein design (Nobel Prize, 2024).

The operational lesson for research programs: licensing and code availability are first-class evaluation criteria, not afterthoughts. A model that cannot run in the required environment is not usable for the program.

Confidence-Calibration Failure Modes

A list of pitfalls that recur in practice:

  1. High pLDDT, wrong orientation. Two domains can each be confidently predicted while the PAE matrix shows inter-domain orientation is undetermined. The cartoon will look definitive; the science isn’t.
  2. High ipTM, wrong interface in cells. A confident complex prediction can correspond to an interaction that does not occur at endogenous expression levels, in the relevant cellular compartment, or in the presence of competitor partners.
  3. Training-distribution overlap. AlphaFold 2 was trained on PDB structures available at the time. Targets with close homologs in the training set are predicted with higher confidence than truly novel targets. Check for near-neighbours in PDB before interpreting a high score on a “novel” target.
  4. Hallucinated ligand poses. AF3-class systems can produce confident-looking ligand placements for chemistry that is out-of-distribution. Treat ligand pose predictions as docking hypotheses, not as bound structures.
  5. Disorder mis-read. Low-pLDDT regions are not “wrong”: they are predictions of disorder. Removing them and re-rendering the structure is a self-deception.
  6. Single state, multi-state biology. GPCRs, kinases, ion channels, allosteric enzymes, and any protein with a meaningful conformational cycle are not described by a single structure, predicted or experimental.

Downstream Applications and Their Limits

Predicted structures feed into:

  • Variant effect prediction. AlphaMissense scored all possible human single-amino-acid substitutions and assigned 57% as likely benign, 32% as likely pathogenic, and 11% as ambiguous (Cheng et al., 2023). Those are model categories, not ACMG/AMP classifications. Clinical use requires validated, framework-compliant interpretation. See Variant Effect Prediction.
  • Protein design. RFdiffusion (Watson et al., 2023) and related systems generate sequences that fold to specified geometries. AlphaFold-predicted structures of the designs are an iteration-loop component, not a substitute for experimental characterization. See Protein Design and Engineering.
  • Ligand and cofactor transfer. AlphaFill enriches AlphaFold models by transferring ligands and cofactors from homologous experimental structures, which is useful for hypothesis generation but not equivalent to a measured bound structure (Hekkelman et al., 2023).
  • Structure-based drug design. Predicted structures support docking and pocket analysis but inherit all the conformational and ligand-chemistry limitations above. See Small Molecule Generation and ADMET.
  • Functional annotation. Predicted folds can suggest functional class by structural homology: useful for orphan proteins, but not a substitute for biochemical or genetic validation.

Open Technical Directions

Two problems still matter:

  • Generative protein modeling at scale. ESM-3 (Hayes et al., 2025), from EvolutionaryScale, is a multi-modal generative protein language model that models sequence, structure, and function in one architecture. The relevance to prediction (as opposed to design) is the convergence: structure-prediction models and design models share a representational substrate. A practical implication is that future “prediction” workflows will increasingly produce ensembles and conditional generations, not single structures.
  • Affinity-aware interaction models. Boltz-2’s affinity head (Passaro et al., 2025, preprint) is an early example of folding-class models reaching into binding-affinity territory. Treat early claims with the same skepticism applied to any structure-then-ranking pipeline: docking-score correlation with experimental affinity is weak across docking studies.

Common Questions

Did AlphaFold “solve protein folding”?

For many single-chain targets at CASP14, AlphaFold 2 achieved accuracy competitive with experimental structures (Jumper et al., 2021). That benchmark result did not solve function, mechanism, conformational dynamics, binding, or every protein class. The capability change is substantial; the framing matters.

Should I still pursue experimental structure determination?

Yes, when the decision requires state-specific geometry, uncertain complexes, ligand-bound poses, water networks, dynamics, or a level of evidence not supplied by prediction. Whether an experimental structure is required for a regulatory submission depends on the product, claim, context of use, and regulator engagement. Predicted structures are research inputs, not automatic substitutes for cryo-EM, X-ray, or NMR.

Is AlphaFold 3 actually better than AlphaFold 2 for proteins-only tasks?

For single-chain protein-only predictions, AlphaFold 2 remains the well-characterized baseline. AlphaFold 3’s value is in the expanded chemical scope (nucleic acids, ligands, ions, and modifications). For a single-chain monomer that does not require that scope, AlphaFold 2 is the simpler, better-understood, more reproducible choice.

How do I cite AlphaFold predictions in a paper?

Methods should cite the underlying paper for the system used (Jumper et al., 2021; Abramson et al., 2024), cite the database paper when using AFDB models (Varadi et al., 2024; Varadi et al., 2022), and report the version and relevant confidence metrics, including the pLDDT distribution and PAE for residue pairs of interest. Predicted structures require the same provenance as any other computational result that informed experimental design.

Can I use AlphaFold 3 commercially?

Read the current licence before use. The AlphaFold 3 terms changed after the May 2024 release. If commercial or on-premise use is essential, compare the current terms and artifacts for separately developed models such as Boltz-1 and Chai-1; do not infer licence compatibility from benchmark similarity.

Is structure prediction the same as protein design?

No. Prediction takes sequence and asks “what does this fold to?” Design takes a target geometry or function and asks “what sequence achieves this?” The systems share representational machinery and design pipelines often use prediction as a quality-control filter, but the validation evidence required is different. See Protein Design and Engineering.

Demonstrated capability

Single-chain structure prediction for many well-folded protein domains is the best-established demonstrated capability. AlphaFold 2, RoseTTAFold, ESMFold, and related systems make predicted structures routine research inputs, and the AlphaFold Protein Structure Database shows the capability at proteome scale. AlphaFold 3, Boltz, and Chai extend the demonstrated scope to biomolecular interactions, but with narrower confidence guarantees and greater dependence on ligand chemistry, interface quality, and access terms.

Theoretical capability

Theoretical claims include dependable ensemble prediction, reliable affinity ranking, routine induced-fit pocket discovery, and structure prediction that selects the correct biological state for a program without extra context. These are plausible directions because current systems already encode useful structural priors, but the evidence is not strong enough to treat static predicted coordinates as mechanism, binding thermodynamics, or state-specific biology.

Beyond current capability

Beyond current capability includes replacing experimental structural biology for mechanism, conformational cycles, ligand-bound geometry that drives chemistry, and water or protonation networks. Protein structure prediction supplies a hypothesis about geometry. It does not solve function, kinetics, free energy, cellular localization, expression, or pharmacology. Regulatory acceptability is context-specific and cannot be inferred from the model name or confidence score.

Evidence that would change the assessment

Structure prediction becomes more promising with blinded benchmarks and prospective experiments that test complexes, ligands, conformational states, and affinity ranking against unseen biology and chemistry. Stronger evidence would include calibrated confidence for ligand poses, independent reproduction of AF3-class open models, and program-level reports showing that predicted structures change chemistry or mutagenesis decisions with measured success. Static benchmark accuracy alone is not enough for those claims.

Implications for research and program decisions

  • Read pLDDT and PAE before interpreting the structural rendering. A visually plausible structure does not supersede its confidence metrics.
  • For consequential decisions, compute multiple predictions (different seeds, MSA depths, or systems) and look at consensus. A prediction that flips under reasonable perturbation is a low-confidence prediction regardless of pLDDT.
  • For drug-discovery work, do not commit chemistry to a predicted pocket without (a) an experimental structure of a homolog or (b) explicit biochemical evidence that the pocket exists. Cryptic pockets are systematically under-represented in predicted structures.
  • For antibody-antigen work, treat any single prediction as a hypothesis-grade interface until validated by mutagenesis, alanine scanning, or epitope mapping.
  • For conformational-cycle proteins (GPCRs, kinases, transporters), predict multiple states (different templates, conformation-specific MSAs) and validate against any available state-specific experimental structures before functional interpretation.
  • License and deployment diligence are part of method selection. Verify current code, weights, terms, dependencies, and intended use for every model.
  • Cite the underlying paper and the model version. “AlphaFold” without a version is ambiguous; 2, 2.3, 3, and Multimer are different systems with different scope and different licenses.

Cross-References