EC-Less Reactions Are a Distinct Annotation-Resolution Barrier

An EC number is an enzyme-classification identifier used to connect a metabolic reaction to candidate gene functions. In the annotation-gap study, reactions without EC numbers formed a distinct class of annotation problem because the available homology and functional-annotation evidence was harder to connect to stoichiometry-defined reactions. ₁

Key Evidence

Among 201 gapfilled enzymatic reaction-organism pairs, 50 (24.9%) lacked an EC number in ModelSEED and were designated “dark reactions.” ₁ Only 8 of these 50 (16%) received candidate assignments, compared with 88 of 151 (58.3%) reactions with known EC numbers. ₁ The study resolved 96 of 201 pairs (47.8%) through an evidence-triangulation pipeline combining gapfilling, Fitness Browser phenotypes, pangenome annotations, GapMind pathway evidence, and BLAST homology. ₁ The unresolved set contained 105 pairs (52.2%), including most of the EC-less reactions. ₁ This pattern supports the interpretation that missing enzyme classification is not merely a cosmetic annotation defect but a measurable barrier to transferring evidence across databases and analytical steps. ₁

The essential-metabolome pilot supports treating pathway evidence as complementary rather than equivalent to reaction-level EC evidence: GapMind predicted 17 of 18 amino-acid biosynthesis pathways in all 7 successfully mapped organisms, but the analysis did not establish the gene-level identity of every pathway step. ₂ This distinction is relevant to EC-less reactions because pathway completeness can indicate that a metabolic capability is present without supplying the missing reaction-to-enzyme identifier. ₂

Why EC-Less Reactions Resist Resolution

The pipeline used EC numbers to connect gapfilled reactions to Fitness Browser gene annotations, alternative Bakta annotations, GapMind pathway evidence, and Swiss-Prot exemplar sequences. ₁ NB03 matched gapfilled reaction EC numbers to Fitness Browser gene annotations through pangenome gene clusters and resolved 51 of 201 pairs (25.4%) with 107 gene candidates. ₁ NB04 queried Bakta annotations for alternative EC numbers and product-name matches, added 22 newly resolved pairs (10.9%), and produced 1,459 Bakta EC candidate entries. ₁ NB06 downloaded 328 Swiss-Prot exemplar sequences for 75 of 84 unique ECs, identified 154 DIAMOND hits, and contributed to the final 96 resolved pairs through evidence triangulation. ₁

For an EC-less reaction, the reaction stoichiometry remains available, but the standard enzyme-classification key used to retrieve homologs and compare annotations is absent. ₁ The study therefore treated the EC-less category as a limitation of evidence linkage rather than as evidence that the underlying reaction lacks a gene in the organism. ₁

The platform’s data structure supports this identifier-linkage interpretation. The KBase Data Lakehouse reaction table has no EC-number column, so no direct EC-to-reaction mapping exists there. ₃ The pitfalls digest instead recommends bridging through eggNOG KEGG_ko / KEGG_Reaction annotations to the KEGG R-numbers in reaction.abbreviation, or using GapMind pathways for pathway-level analysis. ₃ That bridge is itself incomplete: only ~80% of reactions populate reaction.abbreviation (44,904 / 56,012), and the other 20% have no external ID at all. ₃ The sources do not report whether the 50 EC-less gapfilled reactions fall disproportionately within that unlinked fraction, so any such overlap remains an untested hypothesis. _1,3

The AlphaFold multiple sequence alignment (MSA; the set of aligned homologous sequences used for structure prediction) analysis refines this linkage problem by showing that sequence-structural evidence is itself uneven: among 38,051,842 gene cluster–UniProt pairs, MSA depth and domain-hit count had Spearman ρ = 0.7563, with mean domain hits increasing from 0.59 at MSA depth < 10 to 10.83 at MSA depth ≥ 10,000. ₄ Thus, an EC-less reaction may lack its principal reaction-to-enzyme key while its candidate proteins may also occupy poorly represented sequence space; this is a prioritization hypothesis, not a demonstrated property of the 50 EC-less reactions. ₄

Relation to Functional Dark Matter

The EC-less category overlaps operationally with the broader problem of genomic-under-representation, but the two problems are not identical. ₁ EC-less reactions are defined by missing enzyme-classification identifiers in ModelSEED, whereas functional dark matter concerns genes or proteins whose functions remain poorly characterized or unassigned. ₁ The study found that EC-less reactions were difficult to resolve even within a workflow that included BLAST, pangenome conservation, fitness evidence, and pathway evidence. ₁

This finding refines structural-annotation-gap by showing that a missing reaction-to-enzyme identifier can block downstream evidence integration even when sequence and phenotype data exist. ₁ The AlphaFold analysis supports the need to distinguish this identifier-level barrier from a broader structural annotation gap: only 28.7% of the 132,531,501 gene clusters bridged successfully to AlphaFold MSA depths, while 83.8% had at least one InterProScan domain annotation. ₄ It also supports evidence-triangulation-for-functional-annotation: multiple evidence streams improved overall resolution, but their integration remained constrained when reactions lacked a common functional anchor. ₁

Relation to Metabolic-Model Gapfilling

The EC-less barrier arose within a gapfilling workflow in which conditional gapfilling addressed 38 false-negative organism–carbon-source cases and reportedly added 219 reactions, broken down in the report as 201 enzymatic, 14 transport, and 12 exchange. These reported components do not add up to the reported total, so the breakdown is internally inconsistent in the source and should be read with that caveat. ₁ The models used default ModelSEED gapfilling, which minimizes the number of added reactions but does not guarantee biological optimality, and multiple valid gapfilling solutions may exist for a false-negative case. ₁

Consequently, an EC-less reaction combines two uncertainties: uncertainty about whether the selected gapfilled reaction is the biologically correct solution and uncertainty about which gene performs that reaction. ₁ This interaction strengthens the case for treating metabolic-model-gapfilling and EC-less annotation as linked but separable validation problems. ₁

The essential-metabolome analysis supports this separation: its GapMind pathway-completeness calls were computational, used complete or likely_complete categories, and did not establish that complete pathways were essential for viability. ₂ In particular, the apparent Desulfovibrio vulgaris serine gap was a pathway-level hypothesis requiring growth testing, because a divergent or unannotated pathway could have been missed. ₂ Thus, pathway presence can help prioritize reaction and gene candidates, but it cannot by itself resolve an EC-less reaction or validate a gapfilled reconstruction. _2,1

Interpretation and Limits

The evidence for an EC-less resolution barrier is strong within this dataset because resolution was directly compared between 50 EC-less and 151 EC-annotated reaction pairs. ₁ The result should not be interpreted as a universal estimate for all metabolic databases or organisms because the dataset contained 14 organisms, 12 of which were Proteobacteria, and because the models were based on automated RAST annotations with systematic errors. ₁

The AlphaFold comparison adds a separate coverage limitation rather than changing that EC-specific result: its 29.3% bridge coverage was biased toward better-studied organisms, and the remaining 70.7% may contain a larger annotation gap, although that inference was not directly measured. ₄ This supports retaining the dataset-specific qualification above rather than treating structural evidence availability as a universal proxy for EC assignment. ₄

GapMind provided only partial support for resolving these gaps because it covers approximately 80 carbon and amino acid pathways rather than full metabolism, and its available KBase Data Lakehouse output reports pathway-level step counts rather than individual step identities. ₁ The essential-metabolome pilot further illustrates this coverage and mapping limitation: only 7 organisms were mapped successfully from an underlying essential-gene collection of 45, and the Escherichia coli K-12 genome had 0 GapMind predictions in the relevant KBase pangenome collection. ₂ These results refine the interpretation of pathway absence: an uncalled pathway or step may reflect database coverage or detection limits rather than a true metabolic absence. ₂

The study also reported that baseline flux balance analysis (FBA; constraint-based simulation of metabolic growth) achieved 42.5% overall accuracy across 574 organism–carbon-source combinations and produced 330 false positives, indicating that model permissiveness could complicate interpretation of any reaction selected by gapfilling. ₁ The central implication is therefore prioritization rather than abandonment: EC-less reactions should receive dedicated functional-characterization workflows, while their candidate assignments should not be treated as equivalent to assignments supported by established EC-linked evidence. ₁ The report names the 50 EC-less gapfilled reactions as high-priority targets for computational enzyme-prediction tools, such as DeepEC and CLEAN, and for experimental biochemistry. Their enzyme assignments remain to be characterized. ₁

For the related computational pathway evidence, see Essential Metabolome: GapMind Pathway Analysis Across Essential-Gene Organisms.

Open Directions

• Use the 50 EC-less reactions, a subset of the 201 gapfilled enzymatic reaction-organism pairs, to compare alternative stoichiometric gapfilling solutions with gapseq reconstructions, asking which reactions remain stable across reconstruction methods. ₁
• Combine sequence profiles, structure-based searches, and targeted biochemical assays for the unresolved EC-less reaction-organism pairs, asking whether enzyme families can be assigned without an existing EC identifier. ₁
• Reanalyze the 50 EC-less reactions with the 23 available gene-protein-reaction (GPR; the rule linking genes to the reactions their products catalyze) insertions and targeted gene-knockout or CRISPRi experiments, asking whether phenotype changes distinguish competing candidate genes without relying on circular model growth requirements. ₁
• Expand the EC-less analysis from the 14-organism dataset to all 48 Fitness Browser organisms, asking whether the 16% EC-less resolution rate is reproduced across broader phylogenetic and annotation coverage. ₁
• Integrate the 104 GapMind-gapfill pathway pairings with reaction-level annotation and additional pathway databases, asking whether step-level evidence can recover EC-less functions outside the approximately 80 pathways covered by GapMind. ₁
• Run DeepEC and CLEAN on candidate genes for the 50 EC-less gapfilled reactions. Map their predictions through the eggNOG KEGG_Reaction bridge to reaction.abbreviation. This would test whether enzyme prediction recovers reaction links that are missing because the KBase Data Lakehouse has no EC-to-reaction mapping and some reactions have no external ID. _1,3
• Join the unresolved EC-less candidates to AlphaFold MSA depth and InterPro domain counts, asking whether low structural-sequence representation predicts failure of EC-independent functional assignment without conflating that proxy with biochemical validation. ₄
• Extend the pathway test beyond the 7 mapped organisms and validate the apparent D. vulgaris serine gap on serine-free defined medium, asking whether improved coverage and experiment distinguish a true auxotrophy from a missed or divergent pathway. ₂