# How to Classify a Novel Protein Structure into SCOP and CATH: A Step-by-Step Protocol When you have determined a new protein structure, either experimentally or through prediction, you face a practical question: where does this structure sit in the established classification systems? SCOP and CATH are the two major hierarchical classifications of protein domains, and assigning your novel structure to them is a multi-step process that requires careful preparation of input data, informed use of web servers, and critical interpretation of results. This protocol walks through the complete workflow, from preparing your structure file to documenting your final classification, with attention to the known differences between SCOP and CATH that can affect your outcome. ## Understanding What SCOP and CATH Classify SCOP and CATH both organize protein domains into hierarchical levels, but they were built with different philosophies and different protocols. SCOP emphasizes evolutionary relationships and uses manual curation informed by structural comparison, while CATH uses a more automated approach with class, architecture, topology, and homologous superfamily levels. The two systems frequently agree on clear cases, but they can diverge on domain boundaries and on the placement of proteins that sit at the edges of fold families. A systematic comparison of SCOP and CATH found large and unexpected differences between the two classifications with respect to their domain definitions and their hierarchical partitioning of fold space at every level of the two systems. These differences matter because ignoring them leads to problems when the classifications are used to train or benchmark automatic structure comparison methods. For your practical work, this means you should not assume that a SCOP classification and a CATH classification for the same structure will match. You need to record both and understand why they might differ. The same comparison study proposed a consistent mapping between SCOP and CATH and defined a benchmark set that reduced errors made by structure comparison methods such as TM-Align. This mapping and the associated interactive browser are useful tools when you need to reconcile a SCOP assignment with a CATH assignment for the same domain. ## Preparing Your Structure File for Classification Before you submit anything to a classification server, you need a clean structure file. The quality of your input determines the quality of your classification, and poor input files produce misleading results. ### Obtaining a Valid PDB or mmCIF File Your starting point is a coordinate file in PDB or mmCIF format. If your structure is already deposited in the Protein Data Bank, you can retrieve the official file directly. The PDB houses over 242,000 macromolecular structural models alongside much of the experimental data that underpins them, and this archive enables researchers to compare new entries against a vast collection of solved structures. If your structure is not yet deposited, you will need to export coordinates from your refinement or modeling software in a format that the classification servers accept. For predicted structures from AlphaFold or similar tools, you need to be aware that the file format and the quality metrics differ from experimental structures. The AlphaFold Protein Structure Database provides millions of predicted structures, and domain classification resources have begun to incorporate these predictions alongside experimental structures. However, the classification servers for SCOP and CATH were built primarily around experimental structures, and you should check whether the server you plan to use accepts predicted models. ### Checking Structure Quality Before Submission Structural bioinformatics requires understanding the nuances of the underlying experimental data, data encoding conventions, and quality control metrics that can affect a model's precision, fit-to-data, and comparability. Before classification, examine your structure for the following issues: - Missing residues or gaps in the chain that could affect domain boundary detection - Unusual residue numbering that might confuse the server - Multiple chains in the asymmetric unit that need to be separated - Low-resolution experimental data that produced poorly defined regions - For predicted models, low per-residue confidence scores that indicate unreliable regions If your structure has multiple chains, you should typically classify each chain separately. Domain classification operates on individual chains, and submitting a biological assembly with multiple chains can produce confusing results. ### Selecting the Chain and Domain Boundaries SCOP and CATH classify domains, not entire chains. A single chain can contain multiple domains, and the boundary between domains is not always obvious. The CATHEDRAL algorithm was developed specifically to address this problem. It uses an iterative protocol that first locates known folds within a multidomain context using a fast secondary-structure-based method, then applies a residue-based double-dynamic programming algorithm to align members of the target fold groups against the query structure to identify the closest relative and assign domain boundaries. In a benchmark against other publicly available structure comparison methods, CATHEDRAL showed superior performance in fold recognition and alignment accuracy. If a novel multidomain structure contains a known fold, CATHEDRAL locates it in 90 percent of cases with fewer than 1 percent false positives. For nearly 80 percent of assigned domains in a manually validated test set, the boundaries were correctly delineated within a tolerance of ten residues. This performance information is directly useful to you. If you are using CATHEDRAL or a server that implements similar logic, you can expect reliable domain boundary assignment for most structures that contain recognizable folds. But you should also expect that for some cases, previously classified domains are very remotely related to your query chain, and embellishments to the core of the fold cause significant differences in domain sizes. In those cases, manual refinement of the boundaries is necessary. ## The SCOP Classification Workflow SCOP classification requires you to navigate the SCOP database and its associated search tools. The workflow involves submitting your structure for comparison against the SCOP hierarchy and interpreting the results in the context of evolutionary relationships. ### Accessing the SCOP Database and Search Tools The SCOP database is maintained as a curated resource that organizes protein domains according to their evolutionary relationships. To classify a novel structure, you need to access the SCOP search interface and submit your structure for comparison. The search tools typically accept a PDB identifier or a coordinate file. When you submit a structure, the server performs structural comparisons against the existing SCOP entries and returns the closest matches. The results will show you which SCOP superfamilies and folds your structure most closely resembles, along with statistical measures of the similarity. ### Interpreting SCOP Search Results The SCOP hierarchy has the following levels, from most general to most specific: - Class: the secondary structure composition, such as all-alpha, all-beta, alpha/beta - Fold: the arrangement of secondary structure elements in space - Superfamily: proteins that share a common evolutionary origin - Family: proteins with clear sequence and structural similarity When you receive search results, you need to look at the top matches and evaluate whether the similarity is strong enough to support a classification. A high-scoring match to a particular superfamily suggests that your structure belongs to that superfamily, but you need to examine the alignment details to confirm that the match covers the entire domain and is beyond a local similarity. ### Handling Structures with No Clear SCOP Match If your structure does not produce a convincing match to any existing SCOP entry, you have several options. You can classify it as a new fold if the structure is genuinely novel, or you can leave it unclassified if the evidence is insufficient. The decision requires judgment, and you should document the evidence that supports your choice. The known differences between SCOP and CATH mean that a structure with no clear SCOP match might still have a clear CATH match, or vice versa. You should check both systems before concluding that your structure represents a genuinely new fold. ## The CATH Classification Workflow CATH classification follows a similar overall pattern but with different tools and a different hierarchy. The CATH system uses class, architecture, topology, and homologous superfamily levels, and its automated pipeline processes new structures through a series of comparison steps. ### Using the CATH Server for Domain Assignment The CATH server accepts a structure file and runs it through the CATH pipeline. This pipeline includes domain boundary assignment, structural comparison against the CATH database, and assignment to the appropriate hierarchical levels. The CATHEDRAL algorithm is integrated into this pipeline, providing the domain boundary detection and fold recognition capabilities described earlier. When you submit a structure to the CATH server, you should receive results that include the predicted domain boundaries and the closest CATH matches for each domain. The results will typically include confidence scores and alignment statistics that help you evaluate the reliability of the assignment. ### Understanding CATH Hierarchical Levels The CATH levels are defined as follows: - Class: the secondary structure content, similar to SCOP class - Architecture: the orientation of secondary structure elements, regardless of their connectivity - Topology: the connectivity of secondary structure elements, which corresponds roughly to the SCOP fold level - Homologous superfamily: proteins that share a common ancestor, detected through sequence and structural similarity The architecture level is a distinctive feature of CATH. It groups structures based on the overall arrangement of secondary structure elements without requiring specific connectivity, which means that structures with different topologies can share the same architecture. ### Comparing CATH Results with SCOP Results After you have obtained both SCOP and CATH classifications, you need to compare them. The systematic comparison of SCOP and CATH demonstrated that the two hierarchies result from different protocols, which can produce differing classifications of the same protein. You should expect some level of disagreement and be prepared to document it. The comparison study also extracted additional connections in the topology of the protein fold space from the orthogonal features contained in SCOP and CATH. This means that the two systems capture different aspects of protein structure, and using both together gives you a more complete picture than using either alone. ## At a Glance: Classification Decision Table The following table summarizes the key decisions you will make during the classification process and the criteria that should guide each decision. | Decision Point | Primary Criterion | Action When Criterion Is Met | Action When Criterion Is Not Met | |---|---|---|---| | Structure quality | Resolution, completeness, confidence scores | Proceed to domain boundary assignment | Refine structure or note limitations before classification | | Domain boundary assignment | CATHEDRAL or equivalent algorithm confidence | Accept automated boundaries within tolerance | Manually refine boundaries using structural evidence | | SCOP superfamily assignment | High-scoring structural match covering full domain | Record SCOP classification with match statistics | Check CATH for alternative assignment or consider new fold | | CATH homologous superfamily assignment | High-scoring match with consistent domain boundaries | Record CATH classification with match statistics | Check SCOP for alternative assignment or consider new topology | | SCOP-CATH agreement | Consistent classification across both systems | Document agreement and proceed to reporting | Document disagreement and analyze source of divergence | ## Practical Implementation Steps for Classification The following steps provide a concrete workflow that you can follow for each novel structure. These steps assume you have a coordinate file ready and access to the relevant web servers. ### Step 1: Validate Your Structure File Run basic validation on your coordinate file before submission. Check for the following: - The file opens correctly in a structure viewer - The chain you want to classify is complete and properly numbered - The file contains only the chains you intend to classify - For predicted structures, the confidence metrics are recorded and accessible If your structure is deposited in the PDB, verify that the deposited file matches your intended submission. The PDB provides official data resources that you can use to retrieve and verify your structure. ### Step 2: Submit to the CATH Server First Submit your structure to the CATH server first because the CATH pipeline includes automated domain boundary assignment. The CATHEDRAL algorithm will identify domains and assign boundaries, giving you a starting point for understanding the domain organization of your structure. Record the following from the CATH results: - The number of domains identified - The boundaries of each domain - The closest CATH matches for each domain - The confidence scores for each assignment ### Step 3: Submit to the SCOP Server Submit your structure to the SCOP search tools using the domain boundaries you obtained from CATH. If you submit the full chain, the SCOP tools will perform their own domain analysis, which may differ from CATH. If you submit individual domains, you are testing whether each domain has a SCOP match. Record the following from the SCOP results: - The closest SCOP matches for each domain - The SCOP class, fold, superfamily, and family assignments - The alignment statistics for each match ### Step 4: Compare and Reconcile the Results Compare the SCOP and CATH assignments for each domain. Use the SCOP-CATH mapping and benchmark set to understand how the two systems relate for your type of structure. The consistent mapping between SCOP and CATH can be exploited for automated structure comparison and classification, and the interactive browser allows you to explore the relationships between the two systems. If the two systems agree, document the agreement and proceed. If they disagree, analyze the source of the disagreement. Common sources include: - Different domain boundaries assigned by the two systems - Different emphasis on evolutionary versus structural similarity - Genuine ambiguity in the placement of the structure within the fold space ### Step 5: Document Your Classification Create a record that includes the following information: - The structure identifier and source - The classification results from both SCOP and CATH - The domain boundaries used for each classification - The confidence scores and alignment statistics - Any disagreements between the two systems and your interpretation of them - The date of classification and the version of the databases used This documentation is essential for reproducibility. Another researcher should be able to repeat your classification and obtain the same results, or understand why their results differ. ## Records and Measurements for Classification Keeping systematic records of your classification work serves multiple purposes. It allows you to track the reliability of different methods, identify patterns in classification disagreements, and provide evidence for your conclusions in publications or reports. ### Essential Data to Record For each structure you classify, record the following measurements and observations: - The resolution or quality metric of the experimental structure, or the confidence scores for a predicted structure - The number of residues in the chain and the number of domains identified - The domain boundaries assigned by each method - The top match from each classification system with its statistical significance - The final classification at each hierarchical level for both SCOP and CATH - The date and database version used for the classification ### Tracking Classification Confidence Confidence in a classification assignment comes from multiple sources. A high-scoring structural match that covers the entire domain provides strong evidence. A match that covers only part of the domain or that has marginal statistical significance provides weaker evidence. You should record the evidence quality for each assignment. For predicted structures, the confidence in the structure itself affects the confidence in the classification. The ECOD classification of AlphaFold predictions in the Swiss-Prot dataset demonstrated that structure-based classification can identify domains that lack existing sequence-based mappings, but the reliability of these classifications depends on the quality of the predicted structures. ### Maintaining a Classification Log A classification log should include the following columns: | Structure ID | Date | Database Version | Domains | Boundaries | SCOP Assignment | CATH Assignment | Agreement | Confidence | Notes | |---|---|---|---|---|---|---|---|---|---| | Example 1 | 2025-01-15 | SCOP 2.08, CATH 4.3 | 2 | 1-120, 121-250 | alpha/beta, TIM-like | 3.20.20.70 | Yes | High | Clear match | | Example 2 | 2025-02-01 | SCOP 2.08, CATH 4.3 | 1 | 1-180 | all-beta, Ig-like | 2.60.40.10 | Yes | High | Sandwich fold | | Example 3 | 2025-02-10 | SCOP 2.08, CATH 4.3 | 3 | 1-90, 91-200, 201-300 | No match | 1.10.8.10 | No | Low | Boundary ambiguity | This log becomes a valuable reference when you encounter similar structures in the future. It also provides the documentation needed for professional reporting and for responding to questions from collaborators or reviewers. ## Common Failure Patterns in Structure Classification Understanding the ways classification can fail helps you recognize problems early and take corrective action. The following patterns appear frequently in practice. ### Domain Boundary Disagreements The most common source of classification difficulty is disagreement about domain boundaries. SCOP and CATH use different protocols for domain assignment, and the systematic comparison of the two systems found large differences in their domain definitions. A structure that CATH divides into two domains might be treated as a single domain by SCOP, or the boundary between domains might shift by dozens of residues. When you encounter this pattern, examine the structural evidence for the domain boundary. Look for conserved hydrophobic cores, distinct functional sites, and patterns of residue conservation that support one boundary over another. The CATHEDRAL benchmark showed that for nearly 80 percent of assigned domains, boundaries were correctly delineated within a tolerance of ten residues, but the remaining cases required manual refinement. ### Remote Homology and Fold Embellishments Some domains are very remotely related to previously classified domains. The core of the fold is preserved, but embellishments to the core cause significant differences in domain sizes. These cases produce weak matches that may fall below the threshold for confident classification. When you encounter this pattern, you need to decide whether the similarity is sufficient to support a classification. The decision depends on the purpose of your classification. If you are annotating a structure for a database, you may prefer a conservative approach that only assigns classifications with strong support. If you are exploring evolutionary relationships, you may accept weaker matches as evidence of remote homology. ### Predicted Structures with Low Confidence Regions Predicted structures from AlphaFold and similar tools can have regions of low confidence that correspond to disordered segments or domains that are stabilized only in specific contexts. These regions can confuse domain classification algorithms because they lack the well-defined secondary structure that the algorithms use for fold recognition. When you encounter this pattern, you should examine the confidence scores for the regions that affect the classification. If a domain boundary falls in a low-confidence region, the boundary assignment is unreliable. You may need to classify only the high-confidence portions of the structure or note the limitation in your documentation. ### Chimeric or Artificial Constructs If your structure is a chimeric construct or an engineered fusion protein, the domain organization may not reflect a natural evolutionary unit. Classification systems are built around natural protein domains, and artificial constructs can produce misleading results. When you encounter this pattern, you should classify the natural domains separately and document the artificial nature of the construct. The classification of the individual domains is meaningful, but the classification of the chimeric chain as a whole is not. ## Limitations of SCOP and CATH Classification Both SCOP and CATH have known limitations that affect their utility for classifying novel structures. Understanding these limitations helps you interpret your results appropriately and avoid overinterpreting classification assignments. ### Coverage Gaps in the Classifications Neither SCOP nor CATH covers all known protein structures. New structures are deposited continuously, and the classifications are updated on a schedule that lags behind the PDB. Your novel structure may not have a close match in either database simply because the database has not yet incorporated the most recent structures. The PDB continues to grow, and the classification databases must process new entries through their pipelines. The ECOD classification of AlphaFold predictions illustrates the scale of this challenge, with over 542,000 Swiss-Prot protein structure predictions classified into more than 1,032,000 domains. SCOP and CATH face similar scaling challenges, and their coverage may be incomplete for very recent structures. ### Differences in Classification Philosophy The different protocols used by SCOP and CATH produce different classifications for the same protein in a significant number of cases. The systematic comparison of the two systems found differences at every level of the two classifications. These differences are not errors, they reflect different perspectives on protein structure and evolution. SCOP emphasizes evolutionary relationships and uses manual curation to refine automated assignments. CATH uses a more automated pipeline with explicit rules for each hierarchical level. The architecture level in CATH has no direct counterpart in SCOP, and the fold level in SCOP does not correspond exactly to the topology level in CATH. ### Sensitivity to Input Structure Quality Classification results depend on the quality of the input structure. Low-resolution experimental structures or predicted structures with uncertain regions can produce different classifications than high-quality structures of the same protein. The structural bioinformatics guidance emphasizes that quality control metrics affect a model's precision, fit-to-data, and comparability. If your structure has regions of poor quality, you should test whether the classification is robust to variations in those regions. You can do this by trimming the uncertain regions and re-running the classification, or by comparing the classification of your structure with the classification of closely related structures. ## Quality Controls and Verification Steps Implementing quality controls throughout the classification process reduces the risk of incorrect assignments and improves the reliability of your conclusions. ### Cross-Validation with Multiple Methods Do not rely on a single classification method. Use both SCOP and CATH, and if available, use additional classification resources such as ECOD. The ECOD classification provides an evolutionary framework that can complement the SCOP and CATH perspectives, and its integration of predicted structures makes it particularly useful for AlphaFold models. Cross-validation also includes using different structural comparison tools. If you have access to TM-Align or similar tools, you can perform your own comparisons against representative structures from the candidate fold or superfamily. The benchmark set derived from the SCOP-CATH comparison was shown to reduce errors made by structure comparison methods such as TM-Align, so using this benchmark set can improve the reliability of your own comparisons. ### Verification Against Sequence-Based Annotations Structure-based classification should be consistent with sequence-based annotations when those annotations are reliable. If your structure has a clear sequence match to a protein with an established classification, your structure-based classification should agree with that established classification. The ECOD classification of Swiss-Prot predictions found that over 100,000 domains lack existing Pfam mappings, reflecting the extended sensitivity of structure-based classification and identifying domain groups not yet captured by sequence-based profiles. This means that a lack of sequence-based support does not invalidate a structure-based classification, but it does warrant additional scrutiny. ### Documentation of Version and Date Classification databases change over time. A structure that has no match in the current version of SCOP might have a match in a future version, or a classification that seems confident today might be revised as new structures are added. Record the version and date of every database you use. The NCBI provides data resources that are updated on regular schedules, and the EBI provides training and data resources that reflect the current state of bioinformatics databases. Checking the version information for each resource you use ensures that your documentation is accurate. ## Safety and Reproducibility Context While protein structure classification does not involve the physical safety concerns of laboratory work with hazardous materials, it does involve reproducibility concerns that affect the reliability of your research. ### Reproducibility of Classification Results Classification results should be reproducible. Another researcher should be able to take your structure file, run the same classification tools, and obtain the same results. Achieving this reproducibility requires careful documentation of your methods and parameters. The bioinformatics training resources from the EBI and the Galaxy Training Network emphasize the importance of reproducible workflows. The nf-core documentation provides standards for community pipelines that ensure consistent execution. Applying these principles to your classification workflow means documenting every step, from the structure file preparation to the final classification assignment. ### Version Control for Analysis Scripts If you use scripts to process structure files or to analyze classification results, you should use version control to track changes to those scripts. The Carpentries lessons provide foundational training in version control with Git, which allows you to track changes, collaborate with others, and reproduce analyses from any point in the development history. Version control is particularly important when you are developing new analysis methods or adapting existing methods to new types of structures. The ability to return to a previous version of your analysis script can be essential when you discover that a change in the script produced unexpected results. ### Data Management for Classification Records Your classification records are data, and they should be managed with the same care as your experimental data. Store them in a format that is accessible to your collaborators, back them up regularly, and document the metadata that describes how they were created. The Bioconductor project provides documentation on reproducible genomic-analysis workflows that can be adapted to structural bioinformatics. The principles of reproducible research, including clear documentation, version control, and data management, apply equally to structure classification. ## Professional Escalation Criteria Some classification situations require consultation with experts or escalation to more specialized resources. The following criteria indicate when you should seek additional expertise. ### When to Consult a Structural Bioinformatics Specialist You should consult a specialist when you encounter any of the following situations: - Your structure has no convincing match in either SCOP or CATH, and you need to determine whether it represents a new fold - The domain boundaries assigned by different methods disagree substantially, and you cannot resolve the disagreement with structural evidence - Your structure is a predicted model with extensive low-confidence regions, and the classification depends on those regions - You are classifying a large set of structures and need to establish a consistent protocol for handling ambiguous cases A specialist can provide access to additional tools, experience with similar cases, and judgment about the significance of classification differences. ### When to Use Advanced Classification Resources The standard SCOP and CATH web servers may not be sufficient for all cases. Advanced resources include: - The SCOP-CATH mapping browser, which allows you to explore the relationships between the two classifications - ECOD, which provides an evolutionary classification that may capture relationships missed by SCOP and CATH - Specialized domain parsers such as DPAM, which was developed for AlphaFold models The ECOD classification of Swiss-Prot predictions used the DPAM pipeline to classify domains from over 542,000 protein structure predictions, demonstrating the utility of specialized tools for predicted structures. If your work involves large numbers of predicted structures, you should consider whether these advanced resources are appropriate. ### When to Report Classification Uncertainty You should report classification uncertainty in your publications and reports when the classification is not definitive. This includes cases where: - The top match has marginal statistical significance - SCOP and CATH disagree on the classification - The domain boundaries are uncertain - The structure quality is insufficient for confident classification Reporting uncertainty is a professional responsibility. It allows other researchers to understand the limitations of your conclusions and to make their own judgments about the reliability of the classification. ## Building a Decision Framework for Conflicting SCOP and CATH Assignments When SCOP and CATH produce conflicting classifications for the same domain, you need a structured way to decide which assignment to trust or whether to report both. A systematic comparison of the two systems found large and unexpected differences in domain definitions and hierarchical partitioning at every level, so conflicts are not rare anomalies but expected outcomes of the different classification protocols. The following decision framework gives you a repeatable method for resolving these conflicts instead of relying on intuition or preference for one system over the other. ### Step 1: Determine the Source of the Conflict Before you can resolve a disagreement, you must identify whether it originates from domain boundary differences or from hierarchical placement differences. These two sources require different resolution strategies. Domain boundary conflicts occur when SCOP and CATH divide your chain into different numbers of domains or place the boundary between domains at different residue positions. The CATHEDRAL benchmark showed that for nearly 80 percent of assigned domains, boundaries were correctly delineated within a tolerance of ten residues, but the remaining cases required manual refinement. If your boundary difference exceeds ten residues, you should treat the boundary as uncertain and examine the structural evidence directly. Hierarchical placement conflicts occur when both systems agree on the domain boundaries but assign the domain to different folds, superfamilies, or topologies. These conflicts reflect the different emphases of the two systems, with SCOP prioritizing evolutionary relationships and CATH using a more automated protocol with explicit rules for each level. To determine the source, create a simple comparison table for each domain: | Domain | SCOP Boundary | CATH Boundary | Boundary Difference | SCOP Fold | CATH Topology | Conflict Type | |---|---|---|---|---|---|---| | Domain 1 | 1-120 | 1-115 | 5 residues | TIM-like | 3.20.20 | Boundary only | | Domain 2 | 121-250 | 116-260 | 10+ residues | No match | 2.60.40 | Boundary and hierarchy | ### Step 2: Apply the Boundary Resolution Protocol For boundary conflicts, use the following protocol to determine which boundary is more defensible. First, examine the structural features at the disputed boundary region. Look for conserved hydrophobic cores that suggest a compact folding unit, functional sites that should remain within a single domain, and patterns of residue conservation that indicate evolutionary units. The structural bioinformatics guidance emphasizes that understanding the nuances of experimental data and quality control metrics is essential for drawing accurate conclusions from structural data. Second, check whether the disputed region has low structural quality. If your structure has missing residues, poor electron density, or low confidence scores in the boundary region, the boundary assignment is unreliable regardless of which system produced it. For predicted structures, low per-residue confidence scores in the boundary region should trigger additional scrutiny. Third, test the robustness of the boundary by trimming the disputed region and re-running the classification. If the classification of the remaining domain is stable regardless of whether you include or exclude the disputed residues, the boundary difference is unlikely to affect your final assignment. If the classification changes, the boundary difference is consequential and you should report both possibilities. ### Step 3: Apply the Hierarchy Resolution Protocol For hierarchical placement conflicts where both systems agree on boundaries, use the following criteria to evaluate which assignment is more appropriate for your purpose. Consider the strength of the structural match. A high-scoring match that covers the entire domain provides stronger evidence than a marginal match that covers only part of the domain. The consistent mapping between SCOP and CATH developed in the systematic comparison study can help you understand how the two systems relate for your type of structure, and the associated benchmark set was shown to reduce errors made by structure comparison methods such as TM-Align. Consider the evolutionary context. If your protein has clear sequence homologs with established classifications, check whether those homologs support one assignment over the other. The ECOD classification of Swiss-Prot predictions found that over 100,000 domains lack existing Pfam mappings, which means sequence-based support may be absent even for valid structure-based assignments. A lack of sequence support does not invalidate a structure-based classification, but it does warrant additional scrutiny. Consider the purpose of your classification. If you are annotating a structure for a database, you may prefer the assignment with stronger statistical support. If you are exploring evolutionary relationships, you may prefer the assignment that places your domain in a superfamily with known functional or evolutionary significance. ### Step 4: Record the Conflict Resolution Decision For every conflict you resolve, record the following information in your classification log: - The type of conflict (boundary, hierarchy, or both) - The evidence you examined to resolve the conflict - The final assignment you chose and the rationale - The alternative assignment and why you did not choose it - The date and database versions used This documentation serves two purposes. It allows another researcher to understand your reasoning and potentially reach a different conclusion with additional evidence. It also provides a record of how often conflicts arise in your work, which can help you identify systematic biases in either classification system for your particular type of structures. ### Step 5: Apply the Escalation Criteria Some conflicts cannot be resolved with the evidence available to you. The following criteria indicate when you should escalate the conflict to a specialist or use advanced resources. Escalate when the conflict involves a potential new fold. If SCOP assigns your domain to an existing fold but CATH does not, or vice versa, the classification has implications beyond your single structure. A specialist can help you determine whether your structure represents a genuinely new fold or whether the conflict reflects a known difference between the systems. Escalate when the conflict affects a large set of structures. If you are classifying multiple structures and the same conflict pattern appears repeatedly, you need a consistent resolution protocol. A specialist can help you establish criteria that apply across your entire dataset instead of making case-by-case decisions. Escalate when the conflict has functional implications. If your domain is an immunoglobulin-fold domain, for example, the classification affects how you interpret its role in cell-cell recognition, adhesion, or signaling. The IgStrand numbering scheme provides a universal residue numbering approach for immunoglobulin-fold domains that can help resolve classification questions for this important fold family. ### Common Conflict Patterns and Their Resolutions The following patterns appear frequently in practice and have established resolution strategies. The boundary shift pattern occurs when SCOP and CATH agree on the number of domains but place the boundary at different positions. This pattern is common for domains with embellishments to the core fold that cause significant differences in domain sizes. The CATHEDRAL benchmark found that for the remaining cases beyond the 80 percent with correct boundaries, previously classified domains were very remotely related to the query chain, and manual refinement of the boundaries was necessary. Resolve this pattern by examining the structural evidence at the disputed boundary and choosing the boundary that preserves the compact folding unit. The fold versus topology pattern occurs when SCOP assigns a fold that does not correspond to the CATH topology assignment. This pattern reflects the different definitions of these levels in the two systems. The architecture level in CATH has no direct counterpart in SCOP, and the fold level in SCOP does not correspond exactly to the topology level in CATH. Resolve this pattern by documenting the difference and reporting both assignments with an explanation of how the levels relate. The single versus multiple domain pattern occurs when one system divides your chain into multiple domains while the other treats it as a single domain. This pattern is the most consequential because it changes the fundamental unit of classification. Resolve this pattern by examining the structural evidence for independent folding units, including hydrophobic cores, functional sites, and evolutionary conservation. If the evidence is ambiguous, report both possibilities and note the ambiguity in your documentation. ### Integrating the Decision Framework into Your Workflow The decision framework should be applied after you have completed the basic classification workflow for both SCOP and CATH. It is not a replacement for the initial classification steps but rather a structured method for handling the conflicts that those steps reveal. Add the following columns to your classification log to track conflict resolution: | Conflict Type | Evidence Examined | Resolution | Alternative | Rationale | |---|---|---|---|---| | Boundary | Hydrophobic core at 115-125 | CATH boundary | SCOP boundary | Core supports CATH split | | Hierarchy | TM-Align score 0.85 to fold A | SCOP assignment | CATH assignment | Stronger match to SCOP fold | | Both | Low confidence in boundary region | Report both | None | Cannot resolve with current data | This integration ensures that your conflict resolution decisions are systematic, documented, and reproducible. The training resources from the EBI and the Galaxy Training Network emphasize the importance of reproducible workflows, and applying these principles to conflict resolution is consistent with that guidance. ## Frequently Asked Questions ### What is the difference between SCOP and CATH classification? SCOP and CATH are both hierarchical classifications of protein domains, but they use different protocols and emphasize different aspects of protein structure. SCOP emphasizes evolutionary relationships and uses manual curation, while CATH uses a more automated pipeline with explicit rules for each level. The two systems can produce different classifications for the same protein, and a systematic comparison found differences at every level of the two classifications. ### Can I classify a predicted structure from AlphaFold in SCOP and CATH? You can attempt to classify predicted structures, but you need to be aware of the limitations. The SCOP and CATH servers were built primarily around experimental structures, and predicted structures may have regions of low confidence that affect classification. The ECOD classification has incorporated AlphaFold predictions and provides an alternative for predicted structures. You should check whether the specific server you plan to use accepts predicted models. ### How do I determine domain boundaries for a multidomain protein? Domain boundary assignment is a critical step in classification. The CATHEDRAL algorithm provides automated domain boundary assignment and has been benchmarked to locate known folds in novel multidomain structures in 90 percent of cases with fewer than 1 percent false positives. For nearly 80 percent of assigned domains, boundaries were correctly delineated within a tolerance of ten residues. For the remaining cases, manual refinement is necessary. ### What should I do if SCOP and CATH give different classifications? Differences between SCOP and CATH are expected and do not necessarily indicate an error. The two systems use different protocols and can legitimately classify the same protein differently. You should document the disagreement, analyze the source of the divergence, and report both classifications with an explanation of the differences. The SCOP-CATH mapping and benchmark set can help you understand how the two systems relate. ### How do I know if my structure represents a new fold? A new fold classification requires that your structure has no convincing match to any existing fold in either SCOP or CATH. You should exhaust the standard classification resources before concluding that your structure is novel. You should also consider that the classification databases may not yet include all recent structures, and a match might appear in a future database update. ### What quality metrics should I check before classifying a structure? You should check the resolution or quality metric of experimental structures, the confidence scores for predicted structures, the completeness of the chain, and the presence of any unusual features that might affect classification. Structural bioinformatics guidance emphasizes that quality control metrics affect a model's precision, fit-to-data, and comparability, and these factors can influence classification results. ### Can I use ECOD as an alternative to SCOP and CATH? ECOD provides an evolutionary classification of protein domains and has incorporated predicted structures alongside experimental structures. The ECOD classification of Swiss-Prot predictions classified over 1,032,000 domains from more than 542,000 protein structure predictions, demonstrating its scale and utility. ECOD can complement SCOP and CATH, particularly for predicted structures, but it does not replace them. ### How should I document my classification results for publication? Your documentation should include the structure identifier, the database versions used, the domain boundaries, the classification at each hierarchical level for both SCOP and CATH, the confidence scores and alignment statistics, and any disagreements between the systems. This documentation allows other researchers to reproduce your classification and to understand the basis for your conclusions. ## Related Bioinformatics Guides - [Genomic Data Analysis Tools: A Comparative Guide for Researchers](/knowledge/bioinformatics/genomic-data-analysis-tools-a-comparative-guide-for-researchers) - [Oxford Nanopore Sequencing: From Sample to Base Calls](/knowledge/bioinformatics/oxford-nanopore-sequencing-from-sample-to-base-calls) - [Spatial Proteomics Method of the Year: What It Means for Your Research](/knowledge/bioinformatics/spatial-proteomics-method-of-the-year-what-it-means-for-your-research) - [What Is the Monomer of a Protein? Structure & Synthesis](/knowledge/bioinformatics/protein-monomers-amino-acids-peptide-synthesis) - [Mass Spectrometry Protein Identification: From Raw Spectra to Confident Hits](/knowledge/bioinformatics/mass-spectrometry-protein-identification-from-raw-spectra-to-confident-hits) ## References and Further Reading - [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute. - [Bioconductor](https://bioconductor.org/). Bioconductor Project. - [Galaxy Training Network](https://training.galaxyproject.org/). Galaxy Project. - [nf-core Documentation](https://nf-co.re/docs). nf-core. - [The Carpentries Lessons](https://carpentries.org/lessons). The Carpentries. - [Systematic comparison of SCOP and CATH: a new gold standard for protein structure analysis.](https://pubmed.ncbi.nlm.nih.gov/19374763). BMC structural biology, 2009. - [CATHEDRAL: a fast and effective algorithm to predict folds and domain boundaries from multidomain protein structures.](https://pubmed.ncbi.nlm.nih.gov/18052539). PLoS computational biology, 2007. - [ECOD: Classification of domains in AFDB Swiss-Prot structure predictions.](https://doi.org/10.1371/journal.pcbi.1013431). 2026. - [Ten rules for a structural bioinformatic analysis.](https://doi.org/10.1371/journal.pcbi.1013094). 2025. - [IgStrand: A universal residue numbering scheme for the immunoglobulin-fold (Ig-fold) to study Ig-proteomes and Ig-interactomes.](https://doi.org/10.1371/journal.pcbi.1012813). 2025. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.