> ## Documentation Index > Fetch the complete documentation index at: https://docs.boltz.bio/llms.txt > Use this file to discover all available pages before exploring further. # Overview > Guide predictions with structural knowledge to improve accuracy for challenging binding modes. Constraints allow you to incorporate experimental knowledge or structural hypotheses directly into Boltz predictions. While standard predictions to well-defined binding sites work without constraints, they become essential for covalent attachments, cryptic pockets, and ambiguous binding modes. For well-characterized binding interactions (such as reversible inhibitors at orthosteric sites, or known protein-protein interfaces), constraints are optional. The model will find the binding site automatically. *** ## When Should You Use Constraints? Specify specific atom-atom covalent bonds Direct binders to non-obvious or allosteric binding sites Enforce proximity between biomolecular chains or domains Constraints are most valuable when you have prior structural knowledge—SAR data showing a covalent attachment, a crystal structure revealing an allosteric pocket, or biochemical evidence of a biomolecular interaction. *** ## Three Types of Constraints ### Bond Constraints Define **covalent bonds** between specific atoms. | Property | Description | | ------------------- | ------------------------------------------------------------------------- | | **Use Case** | Covalent bonds (acrylamide warheads, nitrile warheads, disulfide bridges) | | **Inputs Required** | Chain + Residue + Atom for both bonding partners | | **Model Behavior** | Hard constraint—these atoms MUST form a bond | Bond constraints are essential for predicting structures with covalent attachments between any biomolecular components—such as covalent inhibitors (e.g., EGFR inhibitors targeting Cys797), disulfide bonds, or peptide cyclization. Step-by-step guide for modeling covalent bonds *** ### Contact Constraints Define **distance restraints** between atoms or residues. | Property | Description | | ------------------- | ---------------------------------------------------------------- | | **Use Case** | Known interaction sites | | **Inputs Required** | Two entities + maximum distance (4-20 Å) | | **Model Behavior** | Predicts a structure using the given information as a constraint | Contact constraints help when you know two regions should interact. They're useful for enforcing proximity between any pair of biomolecules—proteins, ligands, DNA, RNA, or combinations thereof (e.g., molecular glues bridging protein chains, or ligand-DNA contacts). Guide multi-protein assemblies with contact restraints *** ### Pocket Constraints Define **binding site residues** where a ligand/chain should bind. | Property | Description | | ------------------- | ---------------------------------------------------------- | | **Use Case** | Cryptic pockets, allosteric sites, ambiguous binding modes | | **Inputs Required** | Set of residues + which chain should bind | | **Model Behavior** | Directs binder chain toward specified pocket region | Pocket constraints are critical when your target has multiple potential binding sites and you want to focus on a specific one—such as an allosteric site distinct from the orthosteric pocket. The binder chain can also be a polymer, for example, to specify the epitope that an antibody binds to on the antigen. Target specific binding regions with pocket constraints *** ## Using Constraints in Design Projects Constraints become even more powerful in Design Projects, where they guide virtual screening campaigns and iterative design cycles. You can apply the same constraint logic to entire libraries of compounds, ensuring all predictions respect your structural requirements. While constraints can also be used to test multiple hypotheses, pay attention: overly restrictive constraints can force the model into unrealistic conformations. *** ## What's Next? Detailed instructions for configuring each constraint type Apply constraints across virtual screening campaigns