Building a Shared Ontology of Insecticide Modes of Action for Molecular Entomology, Resistance Interpretation, and Cross-Class Innovation

INTRODUCTION

The Insecticide Resistance Action Committee (IRAC) mode-of-action scheme provides a common operational language for grouping insecticidal agents and supporting resistance-management choices. The current IRAC scheme is both a continuously updated resistance-management tool and the product of an industry-led historical classification process [1, 2]. Its utility depends on concise categories that can be communicated across research, product stewardship, and field-decision contexts. That compression is necessary, but it also means that an operational group should not be read as a complete molecular description of every member.

The problem is increasingly visible in discovery. Recent discovery reviews show that new insecticidal chemistry is still organized around target novelty, resistance-breaking potential, selectivity, and deployability rather than around a single mechanistic descriptor [3]. Natural-product-inspired discovery further demonstrates that chemical origin, scaffold class, and biological mechanism are analytically distinct dimensions [4]. A candidate may therefore be chemically novel without acting through a novel target, or it may engage a familiar target through a different site, pose, or physiological sequence.

A shared ontology is needed because current terminology often permits target family, binding process, functional perturbation, intoxication syndrome, and resistance-management grouping to stand in for one another. Four boundaries are foundational: a shared molecular target is not equivalent to identical binding or physiological outcome; target engagement is not equivalent to organism-level efficacy; chemical class is not equivalent to mode of action; and ontology placement is not equivalent to regulatory or resistance-management validation. These are not semantic cautions alone. They determine what evidence must be collected, how contradictory observations are represented, and when an interpretation must remain provisional.

This article therefore proposes an evidence-bearing ontology for insecticide mode-of-action interpretation. It is an original non-empirical synthesis, not a new classification standard, experimental validation, regulatory analysis, or deployment-ready management framework. The ontology is designed to represent entities, typed relations, contextual qualifiers, evidence provenance, uncertainty, and failure-return states across molecular targets, binding, cellular and physiological consequences, resistance, and conditional outcomes. Its central argument is that a useful shared ontology should preserve operational classifications as one view while preventing them from collapsing distinct biological claims.

Why current mode-of-action classifications remain fragmented

Fragmentation begins within nominally shared target groups. The nicotinic acetylcholine receptor (nAChR) family includes multiple subunits, assemblies, species contexts, and ligand-dependent pharmacologies. Within nAChR modulators, compounds assigned to a broad receptor family can differ in subtype engagement, selectivity, metabolism, resistance profile, and toxicity [5, 6]. A family-level label can therefore support broad communication while remaining insufficient to specify the molecular entity engaged, the route from receptor modulation to intoxication, or the expected relationship among compounds in a particular population.

Compound history and organismal context add further discontinuities. Sulfoxaflor illustrates why a shared receptor family cannot substitute for explicit evidence about pharmacology, metabolism, resistance, and cross-resistance [7]. Cross-kingdom target overlap further shows that target identity alone cannot define selectivity, pathway consequence, or operational mode of action [8]. Homologous proteins may participate in different physiological networks, be expressed in different tissues or stages, and be exposed to distinct toxicokinetic conditions. Consequently, neither chemical lineage nor target homology is sufficient to infer equivalent outcomes.

The evidentiary layers themselves are also non-equivalent. Radioligand assays can establish target recognition, but they do not by themselves establish functional perturbation or organism-level efficacy [9]. Binding, channel or enzyme function, cellular signalling, physiological disruption, toxic phenotype, and resistance must therefore be represented as connected but separable claims. Current fragmentation is best interpreted not as the failure of a single scheme, but as the result of different users compressing different evidence dimensions into the same label. The evidence dimensions and interpretive boundaries for current mode-of-action classifications remain fragmented are summarized in Table 1.

Table 1. Why Current Mode-of-Action Classifications Remain Fragmented: Molecular Layers, Targets, Phenotypic Connections, Validation Evidence, Evolutionary Implications, and Interpretive Boundaries

Molecular targets and binding processes

The ontology must first represent targets with greater granularity than a family label. Genomic surveys of Cys-loop ligand-gated ion channels show that target annotation must resolve paralog, subunit, isoform, and species context before pharmacological equivalence can be inferred [10]. A target record should therefore identify the organism, molecular complex, sequence or isoform, tissue and developmental context, and the evidence by which candidacy or engagement was assigned. Genomic presence indicates a possible target entity, not accessibility, binding, functional modulation, or insecticidal efficacy.

Structural evidence adds resolution but does not remove contextual dependence. Structures from pest and non-target RyR systems, together with mapped resistance substitutions, show that sequence context and structural state can materially change how a nominally shared target is interpreted [11–13]. The ryanodine receptor (RyR) evidence connects pest-species structure, beneficial-insect comparison, and mutation-associated functional resistance, yet each study addresses a different evidentiary relation. A predicted site in a recombinant domain is not a confirmed full-length binding pose, structural similarity does not quantify toxicological selectivity, and the effect of a resistance substitution may vary with receptor background and ligand.

Binding events should consequently be stored as typed relations rather than absorbed into the target name. Cryogenic electron-microscopy evidence demonstrates that different diamide families can occupy the same RyR region with distinct poses, while resistance substitutions alter local structure, affinity, and ligand orientation [14]. The ontology should distinguish binding site, pose, affinity, kinetics, reversibility, allostery, experimental construct, and ligand state, and it should link each property to its method and provenance. Direct recognition evidence remains valuable, but it requires functional coupling and biological translation before it can support a complete mode-of-action assertion [9].

Cellular and physiological consequences

After target engagement, the ontology must represent a sequence of potentially causal but separately evidenced transitions: molecular functional perturbation, cellular response, physiological disruption, and organismal phenotype. The spinosyn literature illustrates that mode-of-action attribution matures through iterative alignment of binding, receptor genetics, neuronal physiology, intoxication syndrome, and resistance evidence [15]. The input to each transition should be an evidence-qualified upstream assertion, while its output should be recorded as a distinct downstream observation rather than as automatic confirmation of the preceding mechanism. This supports an evidence-status layer in which relations may be direct, qualified, inferred, contradictory, or unresolved.

The transition from receptor action to phenotype is especially context dependent. Neonicotinoid receptor activation can recruit complex, context-dependent intracellular calcium mechanisms, preventing a one-step inference from receptor label to cellular outcome [16]. Broflanilide exposure shows why downstream endocrine and developmental disruption should be represented as consequences of target perturbation rather than as interchangeable evidence for the primary target [17]. Cellular-response records should therefore carry cell type, dose, timing, life stage, and assay context, while physiological records should describe latency, severity, reversibility, and the evidence connecting them to upstream events. Progression across the proposed chain should remain conditional when functional or temporal linkage is absent.

The proposed synthesis must also accommodate incomplete chains and explicit failure states. Nootkatone provides a useful boundary case in which reproducible toxicological effects coexist with unresolved or potentially multiple molecular mechanisms [18]. High-throughput surface plasmon resonance and molecular simulation can strengthen target-engagement claims, but receptor function and organismal causality remain separate validation steps [19]. Such cases should enter provisional or unresolved states, with return paths to target discovery or causal testing rather than false certainty. Validation would require orthogonal functional assays, temporal or rescue evidence, pathway-specific perturbation, and external tests of whether independent annotators can distinguish primary engagement, downstream response, and conditional efficacy consistently.

Resistance mechanisms and cross-class relationships

Resistance interpretation requires separation of mechanism, evolutionary origin, phenotype, and population context. Resistance ontology must distinguish the biochemical mechanism that changes susceptibility from the evolutionary origin and population context in which that mechanism becomes consequential [20]. A target-site alteration, metabolic pathway, penetration change, sequestration process, or behavioural response should therefore be represented as a determinant–mechanism relation linked to a defined compound, genotype, population, exposure history, and assay. A resistance ratio alone does not identify the mechanism that produced it.

Resistance to Bacillus thuringiensis pesticidal proteins demonstrates that an operational insecticidal class can map to multiple receptor, processing, signalling, and compensatory resistance routes [21]. The codling moth literature similarly shows that cross-class relationships are population-specific composites of target-site, metabolic, penetration, and behavioural mechanisms rather than fixed properties of chemical labels [22]. Cross-resistance should consequently be represented as a directional relation supported by matched susceptibility and mechanistic evidence, because one determinant may affect several compounds asymmetrically while apparently similar resistance phenotypes may arise from co-selected mechanisms.

Mechanistic confidence also varies among resistance assertions. Emerging sequestration and post-transcriptional mechanisms show that a resistance ontology must allow provisional mechanism classes without prematurely treating them as universally established [23]. Gene-cluster knockout provides a stronger causal basis for metabolic-resistance placement than expression association alone, while still retaining compound and genetic-background qualifiers [24]. Ontology entries should therefore distinguish measured from inferred binding changes, functional perturbation from association, and validated resistance determinants from candidates. The evidence dimensions and interpretive boundaries for resistance mechanisms and cross-class relationships are summarized in Table 2.

Table 2. Resistance Mechanisms and Cross-Class Relationships: Molecular Layers, Targets, Phenotypic Connections, Validation Evidence, Evolutionary Implications, and Interpretive Boundaries

Proposed mode-of-action ontology

The proposed ontology begins with a formal entity hierarchy. A shared insecticide ontology should adopt formal definitions, typed relations, evidence provenance, and explicit annotation status rather than relying on label similarity [25]. Its principal entities are compound, chemotype, species-qualified molecular target, target site, binding event, functional molecular perturbation, cellular consequence, physiological or behavioural phenotype, conditional efficacy, and resistance determinant. Stable identifiers and controlled synonyms are required to prevent chemical class, target family, and mechanism from being conflated.

The ontology should then connect those entities through auditable relations such as binds to, modulates, precedes, causes, is associated with, and confers resistance to. Ontology construction should be reproducible and auditable through automated reasoning, quality-control, merging, and release workflows [26]. A maintainable mode-of-action ontology requires modular source files, version control, competency tests, documented releases, and a governed change process [27]. Each relation should carry organism, population, tissue, life stage, dose, route, time, method, provenance, directness, and uncertainty.

Figure 1 presents the hierarchical organization of molecular targets, physiological processes, resistance mechanisms, and insecticidal outcomes within the analytical logic developed in this section.

Alt text

A structured conceptual diagram that presents the hierarchical organization of molecular targets, physiological processes, resistance mechanisms, and insecticidal outcomes, with labelled components, directional relations, contextual modifiers, uncertainty points, and a clear boundary between observed evidence and proposed synthesis.

Graph representation does not itself resolve causal ambiguity. A graph representation is appropriate only if curated observations, mechanistic assertions, and machine-inferred links are stored as different relation and evidence types [28]. Evidence states should include candidate, supported, contradicted, unresolved, and deprecated, with explicit return paths when a compound has a strong phenotype but an unknown target or when molecular findings conflict. Resistance should form a cross-cutting subgraph that links each determinant to the layer it modifies rather than being treated as an attribute of class membership.

The ontology may also support prediction, but inference must remain visibly distinct from observation. Ontology-aware prediction can extend classification to unseen entities, but predicted placement must remain distinguishable from experimentally validated mechanism [29]. Validation would require competency questions, independent expert annotation, inter-rater testing, reconstruction of known cases, counterexample analysis, prospective classification, and decision-use studies performed on a frozen ontology version. The proposed components, evidence bases, boundary conditions, failure modes, and validation requirements are organized in Table 3.

Table 3. Proposed Mode-of-Action Ontology: Components, Evidence Basis, Relations, Boundary Conditions, Failure Modes, and Validation Requirements

Applications in discovery, regulation, and resistance management

For discovery, the ontology can record whether novelty is proposed from phenotype, target hypothesis, direct binding, functional perturbation, or integrated organismal evidence [30]. Discovery portfolios would benefit from separating mechanistic maturity from commercial stage, because candidate novelty and evidentiary completeness do not advance in lockstep [31]. A potent chemotype could therefore remain “mechanism unresolved,” whereas a well-characterized target interaction could remain “efficacy conditional” until biological translation is demonstrated.

An ontology can provide a common evidence layer across target-based, phenotypic, structural, computational, and chemistry-led discovery approaches without treating method novelty as mode-of-action novelty [32]. For evidence organization in regulatory settings, that layer could export consistent definitions, supporting methods, contradictory findings, and boundary conditions. Such an output would be an evidence dossier rather than an approval decision, because jurisdiction-specific standards, benefit–risk assessments, exposure evaluations, and legal requirements remain outside ontology placement.

For resistance management, mixture reasoning should query mechanism independence, resistance frequency, effective dose, joint exposure, and cross-resistance evidence rather than rely only on different class numbers [33]. The same detailed ontology may generate a simplified resistance-management view, but it should return assumptions, population scope, and uncertainty with every output. It cannot establish that a rotation or mixture will be effective under a particular programme without operational and population evidence. The evidence dimensions and interpretive boundaries for applications in discovery regulation and resistance management are summarized in Table 4.

Table 4. Applications in Discovery, Regulation, and Resistance Management: Molecular Layers, Targets, Phenotypic Connections, Validation Evidence, Evolutionary Implications, and Interpretive Boundaries