Abnormal Hearing Phenotypes in “Ignorome” Knockout Mice as Predictors of Cognitive Dysfunction

ABSTRACT

Alzheimer's disease and related dementias affect over 55 million people worldwide and are one of the most pressing public health challenges. Age‐related hearing loss has emerged as a strong predictor of Alzheimer's disease and related dementias risk, raising the possibility that auditory dysfunction may serve as an early biomarker. While the causal nature of the relationship remains uncertain, treating hearing loss, or addressing a shared underlying mechanism, may improve quality of life and slow symptom progression in at‐risk individuals. Current animal models of Alzheimer's disease largely focus on rare familial mutations, limiting their ability to capture the genetic and phenotypic heterogeneity of late‐onset disease. To explore broader genetic contributions and potential links between hearing and cognition, we leveraged data from the International Mouse Phenotyping Consortium, a large‐scale resource that provides standardized phenotyping across thousands of knockout mouse lines. Genes with abnormal auditory phenotypes were more likely to display behavioral abnormalities compared to genes without auditory involvement. Although other sensory modalities such as vision also showed associations with behavioral traits, the links to auditory dysfunction were stronger. Furthermore, higher auditory brainstem response thresholds correlated with the number of behavioral abnormalities across genotypes. Gene Ontology enrichment analyses of genes with auditory and behavioral phenotypes revealed distinct biological processes potentially linking sensory decline and cognitive vulnerability. These findings highlight candidate genes and molecular pathways connecting age‐related hearing loss and Alzheimer's disease and related dementias, provide alternative genetic models that better reflect disease complexity, and suggest new avenues for early detection and intervention.

Using data from the International Mouse Phenotyping Consortium, we analyzed over 9000 knockout mouse lines to test whether sensory impairments predict behavioral abnormalities. Knockouts with abnormal hearing showed higher proportions of behavioral deficits, supporting hearing loss as a potential biomarker of cognitive vulnerability. Created in BioRender. Vicencio, S. (2025) https://BioRender.com/ibix0i5.

Abbreviations

ABR

auditory brainstem response

AD

Alzheimer's disease

ADRD

Alzheimer's disease and related dementias

ARHL

age‐related hearing loss

GO

Gene Ontology

IMPC

International Mouse Phenotyping Consortium

KO

knockout (mouse)

PTA

pure tone average

1. Introduction

Dementia is a clinical syndrome that encompasses multiple diseases impairing cognition, emotion, and everyday functioning. It remains one of the most pressing public health challenges, affecting over 55 million people and generating nearly 10 million new cases each year [1, 2]. Dementia ranks as the seventh leading cause of death and contributes substantially to disability and dependency among older adults [3]. The economic cost of dementia reached $1.3 trillion USD in 2019, with nearly half of this burden falling on informal caregivers [2]. These figures highlight the urgency of identifying modifiable risk factors and early indicators that could delay or mitigate disease onset.

Among these, hearing loss has gained increasing attention. Midlife hearing loss may contribute to ~9% of dementia cases globally [4], and individuals with hearing loss are estimated to have a 94% increased risk of developing dementia [5, 6]. Hearing loss is also common, largely preventable, and treatable, making it a promising candidate for intervention. Despite this, the nature of the relationship between hearing loss and dementia remains unclear. Whether hearing loss directly contributes to cognitive decline, if its impact is indirect via social isolation, or both reflect shared mechanisms is still under debate. Interventions, such as hearing aids, have shown potential to improve quality of life and may help reduce symptom burden or delay onset of cognitive decline, even when causality remains unconfirmed [4, 7]. Recent trials (e.g., ACHIEVE) have provided mixed but encouraging results, suggesting that targeting hearing loss could benefit older adults at risk [8, 9, 10, 11, 12].

Understanding the mechanisms linking hearing loss and dementia requires tools that disentangle genetic contributions from environmental and lifestyle factors. Animal models are critical for this purpose, providing genetic control and enabling longitudinal assessment of phenotypic changes under standardized conditions [13]. While numerous mouse models have been developed for Alzheimer's disease and related dementias (ADRD), especially those focused on amyloid and tau pathology, only a small subset have been evaluated for auditory phenotypes. Notably, studies of APP/PS1 and 5xFAD mice indicate that hearing deficits may precede dementia‐like behaviors, although results vary by strain and background [14, 15, 16]. A tauopathy model, however, does not show accelerated hearing loss [17].

These models are typically based on rare, high‐penetrance mutations associated with early‐onset familial Alzheimer's disease [18, 19]. Many combine multiple familial ADRD mutations in a single strain. While valuable for dissecting pathogenic pathways, they represent only a fraction of dementia cases and fail to capture the broader genetic and phenotypic diversity of late‐onset ADRD [18, 20]. Thus, there is a need to explore alternative models that reflect this complexity and uncover novel contributors to disease risk. To address this gap, we leverage resources from the International Mouse Phenotyping Consortium (IMPC), a large‐scale collaborative effort that systematically characterizes the phenotypic effects of individual gene knockouts in mice [21, 22]. The IMPC includes many genes from the so‐called “ignorome”: genes with little or no prior functional characterization. Studying these under standardized pipelines offers a unique opportunity to uncover unexpected genetic contributors to both hearing loss and cognitive dysfunction. This approach enables more inclusive models of ADRD that move beyond familial mutations toward probing the complex landscape of polygenic risk and sensory aging.

Here, we investigated whether sensory phenotypes—particularly auditory dysfunction—in knockout (KO) mice were associated with cognitive and emotional abnormalities like those in human dementia and mild cognitive impairments. Using data from KO lines in the IMPC (Data Release 23.0), we observed that auditory phenotypes frequently co‐occurred with behavioral abnormalities, including those related to cognition and emotion. While other sensory systems, such as visual and tactile abnormalities, also showed associations with behavioral outcomes, the patterns for auditory dysfunction were more pronounced. Furthermore, the severity of hearing loss, measured by elevated auditory brainstem response thresholds, was associated with the number of behavioral abnormalities observed across genotypes. To further investigate the biological processes underlying these relationships, we conducted Gene Ontology (GO) enrichment analyses on genes associated with auditory phenotypes alone, as well as those that also exhibited cognitive, emotional, or broader behavioral abnormalities.

2. Materials and Methods

2.1. Data Source

We used phenotypic data from the IMPC Data Release 23.0 (April 2025), a standardized and large‐scale mouse genetics platform designed to investigate gene function and its implications for human disease using knockout rodent models [21, 22]. This release includes curated phenotype annotations for 9277 single‐gene KO mouse lines. All data were accessed programmatically through the IMPC application programming interface (API) and are publicly available via the consortium's data portal. For our analyses, we focused on phenotypes assessed in the Early Adult pipeline (9–16 weeks old), which is implemented across IMPC centers using a shared core set of phenotyping tests. While some centers may include a small number of additional assays, the core set of phenotyping tests is consistent across sites. We excluded data from the Late Adult pipeline because it includes fewer types of phenotyping tests, omits auditory assessments entirely, and does not include all knockout lines tested in the Early Adult pipeline.

2.2. Animal Care

All animal procedures were conducted by IMPC partner institutions in accordance with their respective national and institutional ethical guidelines. The welfare of all mice was regularly monitored, and animal care protocols adhered to the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines. Each participating phenotyping center obtained approval from local animal care and use committees or equivalent regulatory authorities. Detailed descriptions of housing, enrichment, and husbandry conditions are publicly available on the IMPC data portal and have been previously described [23].

2.3. Mouse Lines

The knockout mouse lines included in this study were generated by IMPC centers using either nuclease‐mediated genome editing (e.g., CRISPR/Cas9) or, in earlier project phases, gene‐targeted embryonic stem cell approaches. All lines were maintained on a C57BL/6N genetic background, with supporting colonies derived from C57BL/6NJ, C57BL/6NTac, and C57BL/6NCrl. Each knockout targeted a single protein‐coding gene, with phenotyping performed on homozygous mutants when viable, or heterozygotes when homozygous disruption was lethal. Both sexes were tested (typically ≥ 7 males and ≥ 7 females per line) at 9–16 weeks old. Housing and husbandry conditions were standardized across centers as much as possible [23].

2.4. Phenotype Extraction and Classification

We retrieved all phenotype annotations for each gene using the IMPC API, focusing on results from the Early Adult phenotyping pipeline. Phenotypes were classified according to Mammalian Phenotype ontology terms assigned to each knockout line [24, 25], version: releases/2025‐03‐19, obtained from http://purl.obolibrary.org/obo/mp.obo. Using the mp.obo ontology file and the pronto Python package, we mapped Mammalian Phenotype terms to domains relevant to our research questions. Among others, these domains included auditory function, vision/eye physiology, cognitive behavior, emotional behavior, and general behavioral abnormalities. We did not limit our classification to high‐level ontology terms. Instead, we included both broad categories (e.g., “abnormal cognition”) and more specific subterms (e.g., “increased exploration in new environment”) to improve coverage and resolution. For each gene, we generated binary flags indicating the presence or absence of phenotypic abnormalities in each domain. These flags were used for statistical comparisons across sensory and behavioral categories. It is important to highlight that, within the IMPC, Mammalian Phenotype terms can be supported by statistically significant results from multiple independent tests that assess similar aspects of behavior. An MP term is assigned if at least one test shows a significant difference from controls, and agreement across procedures is not required. As a result, the same MP term may be based on different underlying tests across knockout lines.

We quantified auditory dysfunction using average auditory brainstem response (ABR) thresholds, calculated across broadband click and pure tone stimuli (6, 12, 18, 24, and 30 kHz). For each gene, a pure tone average (PTA) was calculated using thresholds at 6, 12, 18, 24, and 30 kHz from the individual animal‐level data available in the IMPC experiment core. This approach follows common practice in audiology research, where averaging thresholds across multiple pure tone frequencies is used to derive a summary measure of hearing sensitivity [26].

2.5. Assessment of Potential Testing Bias Across IMPC Pipelines

Because some behavioral assays are not uniformly implemented across all IMPC phenotyping centers, with some individuals adding different sets of behavioral procedures to the Early Adult pipeline, we evaluated if this could bias the observed associations between sensory and behavioral phenotypes. Using IMPC pipeline annotations, we identified all knockout lines tested within each pipeline and calculated, for each one, the prevalence of abnormal hearing, vision, vibrissae, and behavioral phenotypes, defined as the proportion of tested lines flagged as abnormal within each domain. To avoid spurious results driven by very small sample sizes, we restricted this analysis to pipelines with at least 20 tested knockout lines, which accounted for over 99% of all lines included in the dataset. We then assessed whether pipelines with higher behavioral abnormality prevalence also exhibited systematically higher prevalence of any sensory abnormality using Spearman correlations.

2.6. Gene Ontology Annotation

We performed gene ontology (GO) enrichment analysis using the g:Profiler web tool (version e112_eg59_p19_25aa4782) with g:SCS multiple testing correction method applying significance threshold of 0.05 [27], with the full Mus musculus genome as the statistical background. Analyses were conducted separately for four gene sets: (1) genes with auditory abnormalities, (2) genes with both auditory and other behavioral abnormalities, (3) genes with both auditory and cognitive abnormalities, and (4) genes with both auditory and emotional abnormalities. Enrichment was assessed across the three major GO categories: Biological Process (BP), Molecular Function (MF), and Cellular Component (CC). Significant terms were identified based on adjusted p‐values provided by g:Profiler's built‐in multiple testing correction.

Additionally, for this study we operationally defined ignorome genes as those lacking any experimentally supported GO annotations in the Biological Process or Molecular Function categories. GO annotations for Mus musculus were obtained from the GOA project at the European Bioinformatics Institute. For each gene, we extracted GO annotations together with their associated aspect (Biological Process or Molecular Function) and evidence code. We restricted the analysis to BP and MF annotations supported by direct experimental evidence (EXP, IDA, IPI, IMP, IGI, IEP). Genes were considered to lack experimental functional annotation if they had no BP or MF annotations supported by these evidence codes.

2.7. Candidate Gene Selection Criteria

To identify potential genetic drivers of auditory–behavioral comorbidity, we applied a multi‐step filtering approach to the IMPC dataset. We selected KO strains that exhibited both auditory abnormalities and at least one behavioral phenotype, while excluding those with additional systemic or developmental abnormalities to minimize confounding effects and better isolate gene‐specific links between hearing and other behavior. Specifically, we removed any strains flagged for abnormalities in any of the following top‐level Mammalian Phenotype categories: adipose tissue, cardiovascular system, craniofacial, digestive/alimentary, embryo, endocrine/exocrine glands, growth/size/body region, hematopoietic system, homeostasis/metabolism, immune system, integument, limbs/digits/tail, liver/biliary system, mortality/aging, muscle, pigmentation, renal/urinary, reproductive, respiratory, skeletal, and visual/ocular phenotypes. To ensure the inclusion of genes associated with measurable, but not extreme, auditory dysfunction, we restricted the analysis to KO strains with PTA ABR thresholds between 30 and 65 dB, a range that reflects the moderate hearing loss typically observed in human age‐related hearing loss [28]. Strains with PTA values above this range were excluded to avoid floor effects associated with severe or complete deafness, and those below were excluded to minimize the influence of mild auditory deviations. Then, to summarize the biological relevance of each candidate gene, we retrieved publicly available annotations from the NCBI Gene database (https://www.ncbi.nlm.nih.gov/gene). For each gene, we recorded its reported or predicted function, known tissue expression patterns (with a focus on brain tissues), and human ortholog where available. These annotations were used to construct a descriptive table highlighting potential mechanisms linking auditory and behavioral abnormalities, and to assist in prioritizing candidates for future follow‐up studies.

2.8. Statistical Analysis

We conducted all statistical analyses using Python (version 3.11) and GraphPad Prism (version 10.4.1). As an initial step, we computed Spearman correlation coefficients between binary presence/absence flags for top‐level Mammalian Phenotype categories. Specifically, we tested whether the presence of auditory, vision/eye, or tactile/vibrissae abnormalities was associated with the presence of abnormalities in other domains (e.g., pigmentation, behavior, cardiovascular function). These pairwise correlations were calculated across all genes in the Early Adult dataset.

We also used Fisher's exact test to assess whether genes with sensory abnormalities, specifically auditory, vision, or vibrissae phenotypes, were more likely to exhibit cognitive, emotional, or general behavioral abnormalities. Each sensory category was tested separately, and comparisons were made against the rest of the dataset (i.e., genes without that specific sensory phenotype).

Spearman correlation was also used to assess the relationship between ABR threshold severity and the number of abnormal behavioral phenotypes per gene. This analysis was conducted separately for click thresholds, pure tone average (PTA) thresholds, and high‐frequency PTA (pure tones from 18 kHz and higher), using the subset of genes with available ABR data.

We used Mann–Whitney U tests to compare average ABR thresholds across genes with and without behavioral, cognitive, or emotional abnormalities. These analyses were performed using the full set of genes for which ABR data were available in the Early Adult dataset. Mean threshold differences (in dB) were used as an effect size metric to aid interpretation. All p‐values were corrected for multiple comparisons using the Benjamini–Hochberg procedure.

2.9. Tools, Software, Data and Code Availability

All data retrieval, processing, and statistical analyses were performed using custom Python scripts executed in Visual Studio (version 2022) and GraphPad Prism. Python libraries included pandas for data handling, requests for IMPC API access, scipy.stats for statistical testing, and statsmodels for p‐value correction.

All data analyzed in this study are publicly available through the IMPC data portal. Custom scripts used to retrieve and format the data for the analyses presented here are available from the authors.

3. Results

We began by assessing whether sensory phenotypes, especially auditory, were linked to other behavioral abnormalities across genotypes. To do this, we first computed Spearman correlation coefficients between the presence of abnormal auditory, vision/eye, or vibrissae phenotypes and all other top‐level mammalian phenotypes in the IMPC database. Figure 1 shows the correlation profiles for abnormal hearing phenotypes, and Figure S1 for vision and vibrissa abnormalities. KO mice with abnormal auditory phenotypes (n = 331) showed the strongest statistically significant positive correlation with behavioral (ρ = 0.124, p < 0.0001) and nervous system phenotypes (ρ = 0.104, p < 0.0001). In contrast, vision/eye abnormalities (n = 1988) showed significant correlations with all 23 other top‐level MP categories, the top ones being cardiovascular (ρ = 0.269, p < 0.0001), nervous (ρ = 0.198, p < 0.0001), and growth/size/body region phenotypes (ρ = 0.171, p < 0.0001). Vibrissae abnormalities (n = 124) showed weak but statistically significant correlations with pigmentation (ρ = 0.049), hearing (ρ = 0.047), and behavior phenotypes (ρ = 0.045). While the correlation with integument phenotypes was stronger, it was excluded from interpretation due to the hierarchical nesting of vibrissae phenotypes within the integumentary category. These findings suggest that abnormal hearing phenotypes in KO mice are more specifically and consistently linked to behavioral and nervous system abnormalities, in contrast to visual and vibrissae phenotypes, which show broader or weaker associations across phenotypic domains.

FIGURE 1.

Correlation between abnormal hearing phenotypes and top‐level mammalian phenotypic domains in KO mice. Bar plots show Spearman correlation coefficients between the presence of hearing abnormalities and top‐level phenotype categories defined by the Mammalian Phenotype Ontology. Bar sizes represent the strength of the association (Spearman's rho value), and asterisks (*) indicate statistically significant correlations (p < 0.005).

To assess whether abnormal sensory phenotypes were statistically associated with behavioral abnormalities at the gene level, we performed Fisher's exact tests comparing the co‐occurrence of auditory (Figure 2), visual (Figure S2), and vibrissae phenotypes (Figure S3) with behavioral, cognitive, and emotional abnormalities. Genes associated with abnormal hearing were significantly more likely to also show behavioral (OR = 3.13, p < 0.0001), cognitive (OR = 1.98, p = 0.00025), and emotional abnormalities (OR = 2.11, p < 0.0001) compared to genes without hearing abnormalities. In contrast, genes with visual phenotypes showed weaker associations with behavioral (OR = 1.86, p < 0.0001), cognitive (OR = 1.71, p < 0.0001), and emotional (OR = 1.77, p < 0.0001) domains. Vibrissae phenotypes were significantly associated only with behavioral abnormalities (OR = 2.14, p < 0.0001), with no significant relationship to cognitive or emotional traits. These findings support a stronger and more specific link between auditory dysfunction and behavioral impairments in KO mice than seen with other sensory modalities.

FIGURE 2.

Hearing loss is more common in KO mice with behavioral, cognitive, and emotional abnormalities. The figure shows the proportions of genes with abnormal behavioral (A), emotional (B), or cognitive (C) phenotypes among KO lines with and without hearing abnormalities. (D) shows Odds ratios from Fisher's exact tests comparing the co‐occurrence of hearing abnormalities with behavioral categories. Asterisks indicate statistically significant results (p < 0.001).

After observing that categorical auditory abnormalities were associated with behavioral and nervous system phenotypes, we next asked whether a more fine‐grained, continuous measure of auditory dysfunction, namely, ABR thresholds, would correlate with overall phenotypic burden. Specifically, we calculated the Spearman correlation between ABR thresholds (click, PTA, and high‐frequency PTA) and the total number of abnormal top‐level Mammalian Phenotype terms observed per gene. As shown in Figure 3, the correlations were close to zero for all measures (A) click: rho = −0.0129; (B) PTA: rho = 0.027, and (C) high‐frequency PTA: rho = 0.0393, suggesting no biologically meaningful relationship between type of hearing loss severity and the overall number of phenotypic abnormalities.

FIGURE 3.

Correlation between ABR threshold and overall phenotypic burden. Scatter plots show Spearman correlations between average ABR thresholds (A) click; (B) PTA; and (C) high‐frequency PTA, and the total number of abnormal top‐level Mammalian Phenotype abnormalities present per gene. Each point represents a gene with both ABR data and phenotype annotations. Correlation coefficients were close to zero (click: Rho = −0.0129; PTA: Rho = 0.0027; high‐frequency PTA = 0.0393), indicating no meaningful association between auditory threshold and overall phenotypic disruption.

Since no biologically meaningful correlation was found between ABR threshold and the overall number of abnormal top‐level Mammalian Phenotype terms, we next examined whether auditory thresholds differed specifically across individual Mammalian Phenotype categories. To do so, we grouped KO strains by top‐level Mammalian Phenotype term and compared their ABR thresholds using the Mann–Whitney U test. To evaluate whether hearing loss severity differed across top‐level Mammalian Phenotype terms, we analyzed (A) click‐evoked, (B) PTA, (C) and high‐frequency PTA ABR thresholds in KO strains with abnormalities in each category (Figure 4).

FIGURE 4.

ABR threshold shifts across top‐level Mammalian Phenotype categories. Bar plots show average ABR threshold differences (ΔdB) in knockout strains with abnormalities in each top‐level Mammalian Phenotype category, in (A) click; (B) PTA; and (C) high‐frequency PTA. Each bar represents the average threshold shift for combined sexes (dark bars) and separately for males (red bar) and females (light blue). Only the results of phenotype categories that had significant results (p < 0.001) are shown.

As expected, across all measures, the largest threshold elevations were observed in strains flagged for hearing/vestibular phenotypes (click: ΔdB = 13.46, p < 0.0001; PTA: ΔdB = 8.79, p < 0.0001; high‐frequency PTA: ΔdB = 9.41, p < 0.0001). In addition to auditory‐related phenotypes, behavioral and homeostasis/metabolism abnormalities showed consistent threshold elevations in all three conditions. Behavioral phenotypes were associated with significant increases in ABR thresholds across click, PTA, and high‐frequency PTA stimuli (click: ΔdB = 4.31; PTA: ΔdB = 5.28; high‐frequency PTA: ΔdB = 3.57; all p < 0.0001), and were the only non‐auditory category with consistent effects across both stimulus types and sexes. Homeostasis/metabolism abnormalities also showed smaller but significant threshold elevations across all stimulus types (click: ΔdB = 3.36; PTA: ΔdB = 2.32; high‐frequency PTA: ΔdB = 1.31; all p < 0.0001), although these effects were not consistent across sexes. In contrast, other categories exhibited stimulus‐specific effects. For click‐evoked responses (Figure 4A), muscle system abnormalities were associated with a significant threshold increase (ΔdB = 8.38, p < 0.0001), followed by adipose tissue (ΔdB = 3.21), skeleton (ΔdB = 2.67), limbs/digits/tail (ΔdB = 2.30), and hematopoietic system (ΔdB = 1.68), all significant with p < 0.0001. Pigmentation abnormalities, in contrast, were linked to threshold elevations only in tone‐based measures, including PTA (ΔdB = 7.89, p < 0.0001; Figure 4B) and high‐frequency PTA (ΔdB = 4.75, p < 0.0001; Figure 4C). While elevated thresholds in strains with auditory phenotypes were expected, the consistent association between behavioral abnormalities and increased ABR thresholds across all stimulus types and sexes highlights a potentially important link between auditory dysfunction and neural or cognitive domains.

Thus, we next focused specifically on the behavioral domain. For each gene, we calculated the number of abnormal behavioral Mammalian Phenotype terms and determined if this count was associated with ABR thresholds. As shown in Figure 5, we observed a modest but significant positive correlation: genes associated with higher ABR thresholds also tended to display more abnormal behavioral phenotypes, where (A) click: rho = 0.129, p < 0.0001; (B) PTA: rho = 0.210, p < 0.0001; and (C) high‐frequency PTA: rho = 0.181, p < 0.0001. Unlike the nearly flat correlations with the total Mammalian Phenotype abnormality count, this result supports a more specific link between the severity of auditory dysfunction and the extent of behavioral disruption in knockout mice.

FIGURE 5.

Correlation between ABR threshold severity and number of abnormal behavioral phenotypes. Scatter plots show the relationship between average ABR thresholds (A) click; (B) PTA; and (C) high‐frequency PTA, and the number of abnormal behavioral Mammalian Phenotype terms per gene. Each point represents a gene with both ABR and behavioral data. A modest but positive correlation was observed for all measures (click rho = 0.129, < 0.0001; PTA: Rho = 0.209, < 0.0001; High‐frequency PTA rho = 0.181, < 0.0001), indicating that genes associated with more behavioral abnormalities tend to have higher hearing thresholds.

We also examined whether auditory thresholds differed across specific behavioral Mammalian Phenotype terms using the Mann–Whitney U test. As shown in Figure 6, several behavioral abnormalities were significantly associated with increased ABR thresholds, although the strength and specificity of these associations varied by stimulus type. A group of phenotypes showed consistent threshold elevations across all three ABR conditions, including head bobbing (click ΔdB = 20.31; PTA ΔdB = 23.20; high‐frequency PTA ΔdB = 19.84), abnormal startle reflex (click ΔdB = 18.75; PTA ΔdB = 10.72; high‐frequency PTA ΔdB = 8.33), impaired righting response (click ΔdB = 10.00; PTA ΔdB = 8.50; high‐frequency PTA ΔdB = 6.56), limb grasping (click ΔdB = 10.00; PTA ΔdB = 7.15; high‐frequency PTA ΔdB = 3.30), abnormal locomotor behavior (click ΔdB = 11.25; PTA ΔdB = 6.77; high‐frequency PTA ΔdB = 4.27), decreased startle reflex (click ΔdB = 10.36; PTA ΔdB = 9.62; high‐frequency PTA ΔdB = 11.50), decreased grip strength (click ΔdB = 5.00; PTA ΔdB = 6.50; high‐frequency PTA ΔdB = 4.44), and decreased locomotor activity (click ΔdB = 4.75; PTA ΔdB = 0.98; high‐frequency PTA ΔdB = 1.61). All these associations were statistically significant at p < 0.001 (Figure 6A–C).

FIGURE 6.

ABR threshold shifts are associated with individual behavioral Mammalian Phenotype abnormalities. Bar plots show average ABR threshold differences (ΔdB) for KO strains with each behavioral phenotype, for (A) click; (B) PTA; and (C) high‐frequency PTA.

One phenotype, abnormal gait, showed significant elevations in both click (ΔdB = 11.67) and PTA thresholds (ΔdB = 2.50), but not in high‐frequency PTA (Figure 6A,B). Another group of phenotypes showed significant associations with PTA and high‐frequency PTA thresholds but not with clicks. These included hyperactivity (PTA ΔdB = 6.87; high‐frequency PTA ΔdB = 4.92), abnormal sleep behavior (PTA ΔdB = 8.67; high‐frequency PTA ΔdB = 7.36), abnormal vocalizations (PTA ΔdB = 6.94; high‐frequency PTA ΔdB = 3.75), increased vertical activity (PTA ΔdB = 6.78; high‐frequency PTA ΔdB = 4.46), increased exploration in a new environment (PTA ΔdB = 5.91; high‐frequency PTA ΔdB = 4.17), and decreased thigmotaxis (PTA ΔdB = 4.82; high‐frequency PTA ΔdB = 3.22). All associations in this group were also significant at p < 0.001 (Figure 6B,C).

Several additional phenotypes were associated only with click ABR threshold elevations (Figure 6A). These included tremors (ΔdB = 11.25), increased thigmotaxis (ΔdB = 10.17), increased startle reflex (ΔdB = 6.25), increased grip strength (ΔdB = 3.01), decreased vertical activity (ΔdB = 8.68), decreased exploration in a new environment (ΔdB = 2.00), absent pinna reflex (ΔdB = 7.50), and trunk curl (ΔdB = 9.25). All click‐only effects were significant at p < 0.001.

Finally, two phenotypes showed significant effects in only one tone‐based condition. Impaired pupillary reflex was significantly associated with PTA thresholds (ΔdB = 7.35), while decreased coping response was uniquely associated with high‐frequency PTA thresholds (ΔdB = 9.79). These associations, too, were significant at p < 0.001 (Figure 6B,C).

Together, these results highlight that ABR threshold elevations are not uniformly associated with all behavioral abnormalities, but rather cluster around specific domains related to motor function, reflex modulation, and arousal. Phenotypes such as head bobbing, abnormal startle reflex, and impaired righting response not only appeared consistently across stimulus types but also exhibited some of the highest threshold shifts, emphasizing their potential relevance as behavioral indicators of underlying auditory dysfunction.

Next, to explore the biological foundations of the phenotypic patterns observed, we performed Gene Ontology Enrichment Analysis on the sets of genes associated with (i) auditory, (ii) auditory and other behavioral, (iii) auditory and cognitive behavior, and (iv) auditory and emotional behavior abnormalities (Figure 7). By identifying overrepresented functional categories, we aimed to determine whether auditory and behavioral abnormalities in KO strains shared molecular functions, cellular components, or biological processes. Among the genes associated with auditory abnormalities, in the cellular component domain, the most strongly enriched term was cytoplasm, with 67% of annotated genes and a highly significant adjusted p‐value p < 0.0001. In this set of genes, within the molecular function domain, protein binding was significantly enriched (60% of the genes; p < 0.0001). Several biological process terms were also enriched, including cellular localization (26% of genes; p < 0.0001), sensory perception of sound (6% annotated genes; p < 0.0001), and inner ear receptor cell development (3% of the genes; p < 0.0001). These results confirm functional clustering of auditory‐implicated genes within known sensory pathways and cellular compartments relevant to hearing.

FIGURE 7.

Gene Ontology enrichment analysis of auditory and behavioral associated genes. The figure shows significantly enriched Gene Ontology terms across Biological Process (BP), Molecular Function (MF), and Cellular Component (CC) domains. The x‐axis represents Gene Ontology categories, the y‐axis shows –log₁₀ of the adjusted p‐value, and dot size indicates the number of genes intersecting within each term. In (A) the figure represents all genes with auditory abnormalities, in (B) genes with both auditory and behavioral abnormalities, in (C) genes with both auditory and emotional abnormalities, and (D) genes with both auditory and cognitive abnormalities. Only driver terms were plotted.

Among the genes associated with both auditory and general behavioral abnormalities, Gene Ontology Enrichment Analysis revealed a similar but more specific functional profile. Within the cellular component domain, cytoplasm remained the top enriched term, also with 67% of intersecting genes and an adjusted p‐value p < 0.0001. In the molecular function domain, protein binding was again highly enriched (63 of the gene list; p < 0.0001). Biological process terms with strong enrichment included biological regulation (64% of the genes; p = 0.00481), regulation of cellular component organization (19% of the genes; p = 0.01350), sensory perception of sound (8% of the genes; p < 0.0001), inner ear receptor cell differentiation (5% of the listed genes; p < 0.0001), and cellular localization (27% of the genes; p < 0.0001). These results suggest that genes jointly linked to auditory and behavioral dysfunction converge on functional categories relevant to both sensory processing and cellular organization.

In the gene set associated with both auditory and emotional abnormalities, enrichment results revealed a more selective set of driver terms. The top cellular component was cytoplasm (67% of the genes; p = 0.00728), followed by neuron projection (25% of the genes; p < 0.0001), highlighting the relevance of neural architecture in this subgroup. Enriched biological processes included negative regulation of biological process (48% of the genes; p = 2.28e‐03) and regulation of biological quality (33% of the genes; p = 4.64e‐03). Within the molecular function category, alpha‐L‐arabinofuranosidase activity was nominally enriched (3% of the genes; p = 0.00239), though with a small intersection size, and protein binding remained enriched (67% of the genes; p = 0.01166).

For the gene set associated with both auditory and cognitive abnormalities, enrichment analysis yielded a small number of significant Gene Ontology terms, reflecting a narrower functional focus. In the biological process domain, negative regulation of potassium ion transport was enriched (9% of the genes; adjusted p = 0.0107), pointing toward roles in membrane excitability. Two cellular component terms reached significance: transporter complex (15% of the genes; p = 0.0286) and excitatory synapse (9% of the genes; p = 0.0301), both of which align with known pathways in synaptic signaling and neuronal communication. Additionally, transporter activity was enriched in the molecular function domain (24% of the genes; p = 0.0448). Although fewer terms met statistical thresholds in this group, the identified categories suggest specific molecular mechanisms linking auditory dysfunction with cognitive impairment.

Finally, to identify candidate genes for modeling the link between hearing loss and cognitive decline, we applied a targeted filtering strategy to the IMPC dataset. The goal was to identify genes linking auditory and general behavioral abnormalities in the absence of widespread physiological issues that could act as confounding factors. We selected KO strains that met three criteria: (1) presence of both auditory and behavioral abnormalities, (2) PTA ABR thresholds in the moderate range (30–65 dB), and (3) absence of additional top‐level Mammalian Phenotype abnormalities across major physiological systems. This selection yielded a focused list of 19 genes with selective dual‐domain impairments, minimizing confounding systemic phenotypes. These genes are listed in Table 1.

TABLE 1.

Candidate genes with their known functions and behavioral Mammalian Phenotype.

Gene	PTA (dB)	NS expression (embryo)	NS expression (adult)	Biological function (mice)	Behavioral abnormalities	Human ortholog/diseases
Fuca1 a	60.38	CNS, whole brain	Cerebellum, cortex, frontal lobe	Predicted to be involved in the fucose metabolic process and glycoside catabolic process.	Abnormal learning, neophobia.	FUCA1, implicated in fucosidosis.
Pcp4l1 a	57.57	CNS, whole brain	Cerebellum, cortex, frontal lobe	Not specified.	Decreased locomotor activity.	PCP4L1
Rfxap	56.75	CNS, whole brain	Cerebellum, cortex, frontal lobe	Enables DNA binding activity.	Abnormal voluntary movement, hyperactive.	RFXAP, Immunodeficiency by Defective Expression of MHC Class II
Phf3 a	54.20	CNS, whole brain	Cerebellum, cortex, frontal lobe	Predicted to be involved in regulation of transcription by RNA polymerase II.	Abnormal startle reflex.	PHF3
Iffo2 a	51.75	CNS, whole brain	Cerebellum, cortex, frontal lobe	Not specified.	Increased startle reflex.	IFFO2
E130114P18Rik a	49.25	Predicted gene; expression unknown.	Predicted gene; expression unknown.	Predicted gene; function unknown.	Abnormal voluntary movement, hyperactive.	Not specified
Xpnpep3 a	48.22	CNS, whole brain	Cerebellum, cortex, frontal lobe	Predicted to be involved in proteolysis.	Decreased grip strength.	XPNPEP3, Nephronophthisis‐Like Nephropathy 1
Zfp74 a	47.38	CNS, whole brain	Cerebellum, cortex, frontal lobe	Predicted to be involved in regulation of transcription by RNA polymerase II.	Increased grip strength.	ZNF569
Bicd1	47.00	CNS, whole brain	Cerebellum, cortex, frontal lobe	Enables proteinase‐activated receptor binding activity.	Decreased grip strength, decreased prepulse inhibition, decreased startle reflex.	BICD1
Cbln3	46.25	Not expressed	Cerebellum	Predicted to be involved in maintenance of synapse structure.	Abnormal sleep behavior	CBLN3
Eif1b a	46.00	CNS, whole brain	Cerebellum, cortex, frontal lobe	Predicted to be a translation initiation factor.	Abnormal voluntary movement, hyperactive.	EIF1B
Rell1 a	39.25	CNS, whole brain	Cortex, frontal lobe	Predicted to be involved in positive regulation of the p38 MAPK cascade.	Abnormal voluntary movement, hyperactive. Decreased prepulse inhibition, increased startle reflex.	RELL1
Il1r2	39.00	CNS, whole brain	Cerebellum, cortex	Enables interleukin‐1 binding and interleukin‐1 receptor activity.	Increased grip strength.	IL1R2, Aggressive periodontitis, esophageal cancer, lung cancer, and organ system cancer
Depdc1a a	37.88	CNS	Not expressed	Predicted to enable GTPase activator activity. Predicted to be involved in negative regulation of DNA‐templated transcription and to be part of a transcription repressor complex.	Abnormal learning, increased exploration in new environment.	DEPDC1
Iqgap2 a	37.63	CNS, whole brain	Cerebellum, cortex, frontal lobe	Predicted to be involved in Actin filament organization, regulation of Actin cytoskeleton organization, and the thrombin‐activated receptor signaling pathway.	Increased startle reflex.	IQGAP2
Gpr152 a	36.33	CNS, whole brain	Cerebellum, cortex, frontal lobe	Predicted to be involved in G protein‐coupled receptor signaling pathway.	Absent pinna reflex, decreased grip strength.	GPR152
Gm5447 a	34.63	Predicted gene; expression unknown	Predicted gene; expression unknown.	Predicted gene; function unknown.	Abnormal voluntary movement, hyperactive.	No confirmed human ortholog
Car3 a	32.25	Not expressed.	Not expressed.	Enables nickel cation binding activity.	Impaired righting response.	CA3
Btbd10	31.50	CNS, whole brain	Cerebellum, cortex, frontal lobe	Involved in positive regulation of phosphatidylinositol 3‐kinase/protein kinase B (PI3K/AKT) signal transduction.	Increased vertical activity.	BTBD10

^a Genes that fit our Ignorome criteria.

4. Discussion

Hearing loss is a potentially modifiable risk factor for ADRD, yet the mechanisms linking it to cognitive decline remain unresolved. Studies show that hearing impairment increases dementia risk, but whether this reflects shared biological pathways, causal interactions, or both is unclear. Understanding these connections is important because hearing interventions are more accessible than dementia treatments. To explore genetic foundations of this association, we analyzed the IMPC database, which provides data from thousands of KO mouse lines, many targeting genes with poorly characterized functions. We examined relationships between auditory dysfunction and behavioral abnormalities, focusing on cognitive and emotional traits. We found that KO lines with abnormal hearing were more likely to display behavioral abnormalities than those with normal auditory function or with other sensory abnormalities. This suggests that auditory phenotypes can serve as genetic indicators of cognitive and emotional dysfunction. Notably, when we examined the level of experimental functional characterization of genes phenotyped by the IMPC, we found that roughly one third lacked any experimentally supported Gene Ontology Biological Process or Molecular Function annotations (∼38% across all phenotyped genes and ∼34% among genes associated with abnormal hearing phenotypes). This indicates that many genes implicated in abnormal hearing and related behavioral phenotypes have yet to be studied in detail at the functional level.

Because behavioral assays are not uniformly implemented across IMPC pipelines, we assessed whether pipeline testing differences could account for the observed association between hearing abnormalities and behavioral phenotypes. Behavioral abnormality prevalence did not correlate with the prevalence of hearing (r = 0.17, p = 0.49), vision (r = 0.32, p = 0.17), or vibrissae abnormalities (r = 0.36, p = 0.12) across pipelines. Thus, these findings suggest that differences in behavioral testing are unlikely to explain the stronger association observed between hearing and behavioral/cognitive/emotional phenotypes.

4.1. Association of Sensory and Behavioral Phenotypes

Although auditory, visual, and vibrissae phenotypes all correlated with behavioral abnormalities, the strength and specificity varied. We included multiple modalities to test whether the link with cognition was specific to hearing. Vibrissae phenotypes, representing tactile sensation, were significantly associated only with behavioral abnormalities (OR = 2.14) but showed no relationship with cognitive or emotional traits. Vision‐related abnormalities were associated with behavioral, cognitive, and emotional deficits (ORs 1.71–1.86). Auditory phenotypes, defined by abnormal ABRs, showed stronger associations, particularly for general behavioral abnormalities (OR > 3.2). This suggests that while sensory systems relate to behavioral outcomes, auditory dysfunction may be a more direct indicator. The behavioral associations in the visual domain are likely influenced by the heterogeneity of the “vision/eye” category, which includes diverse traits spanning eye morphology, retinal function, and other anatomical features. The IMPC does not include an objective physiological measure of vision, such as electroretinography, relying instead on this broad category. In contrast, the auditory phenotype was based on a single, well‐defined physiological measure (ABR), making it less susceptible to categorical dilution. This likely contributed to the greater effect sizes observed for hearing‐related associations. Finally, vibrissae phenotypes were underrepresented in the dataset, and the limited number of vibrissae‐related knockouts may have influenced the strength and scope of these findings.

4.2. Hearing Loss Severity Predicts Extent of Behavioral Abnormalities

Across KO lines, higher ABR thresholds correlated with greater numbers of cognitive and emotional phenotype abnormalities, like motor and startle reflex impairments. However, it is important to note that several of the strongest associations involved reflexive responses (e.g., startle and pinna reflexes) that are elicited by acoustic stimuli and therefore depend directly on auditory sensitivity and are probably not reflecting higher‐order cognitive or emotional processes. Interestingly, some abnormalities showed stimulus‐specific patterns: PTA thresholds were linked to abnormal sleep, vocalizations, and exploration, while click thresholds were linked to startle and pinna reflex changes. These distinctions suggest that different aspects of auditory dysfunction may map onto different behavioral outcomes, underscoring the value of examining both auditory measures in parallel.

4.3. Shared Biological Pathways Between Auditory and Cognitive‐Emotional Dysfunction

Genes with auditory phenotypes alone were enriched for cytoplasmic components, protein binding, and hearing‐related processes such as sensory perception of sound and inner ear receptor development. Genes that showed both auditory and general behavioral abnormalities showed enrichment for broader functions, such as cellular localization. Emotional abnormalities were linked to neuron projection components and regulation of biological quality, while cognitive abnormalities were associated with transporter activity, synaptic signaling, and potassium ion regulation.

Potassium ion transport is noteworthy, as it is implicated in both deafness and age‐related hearing loss. KCNQ4 (KV7.4) supports outer hair cell repolarization and is tied to progressive and age‐related loss [29, 30], KCNQ1 (KV7.1) maintains the endocochlear potential, and its age‐related decline may be linked to impaired potassium recycling [31]; and KCNC1 (KV3.1) enables high‐frequency firing in brainstem circuits, with downregulation contributing to auditory temporal deficits [32]. These findings highlight potassium ion transport as a plausible shared pathway connecting sensory decline and cognitive vulnerability.

4.4. Prioritizing Genes for Follow‐Up Studies

While enrichment analyses revealed broad pathways linking auditory and behavioral traits, they did not identify specific genes for follow‐up. To address this, we selected five genes for which their KO strain showed clear associations with both auditory dysfunction and cognitive‐emotional abnormalities, have moderate threshold increases (30–65 dB), and no systemic or motor confounds: Fuca1, Phf3, Rell1, Depdc1a, and Iqgap2. These genes show brain expression and behavioral phenotypes related to learning, memory, exploration, or emotion. Notably, all genes prioritized here for follow‐up met our experimental ignorome criteria, indicating that their functional roles remain largely uncharacterized in experimental mouse studies.

Although not well studied in hearing or cognition, they each have potential or emerging roles that may relate to sensory–cognitive processes. Fuca1 participates in fucose metabolism, with disrupted fucosylation linked to neurodegeneration [33]. Phf3 regulates transcription via RNA polymerase II, potentially influencing neuronal gene expression [34]. Rell1 activates p38 MAPK signaling, central to stress responses, inflammation, and apoptosis [35]. Depdc1a is proposed to act as a transcriptional repressor, and although its role in the brain is unclear, it appeared in both cognitive and emotional phenotype categories in the IMPC database. Iqgap2 regulates cytoskeletal signaling and blood–brain barrier immune responses, with possible roles in plasticity and neuroimmune interactions [36].

Together, these genes converge on immune and stress‐response pathways, increasingly implicated in auditory decline. Chronic inflammation contributes to cochlear and central auditory dysfunction [37], suggesting that genes such as Iqgap2 and Rell1 may link sensory and cognitive outcomes through neuroimmune mechanisms.

4.5. Limitations

Although KO models and the IMPC pipeline provide standardized, large‐scale phenotyping, they cannot capture the polygenic, environmental, and lifestyle complexities of ADRD. The IMPC pipeline is designed for consistency and output but is relatively limited in scope. Behavioral assays prioritize activity, reflexes, and coordination, offering limited coverage of higher cognitive or emotional impairments. Also, another limitation of large‐scale IMPC analyses is that a single behavioral Mammalian Phenotype term can be supported by different underlying tests across knockout lines. For this reason, we think our results should be viewed as an initial screening to highlight candidate associations, rather than definitive conclusions about specific behaviors. Moreover, mice are tested at 9–15 weeks, leaving late‐life decline unassessed, and ABRs are restricted to quiet thresholds, which may miss suprathreshold or noisy‐environment deficits. No ABR data are collected in older cohorts, preventing evaluation of progression with age.

Another key limitation is genetic background: all lines are on C57BL/6N, which carries a Cdh23 mutation causing early high‐frequency hearing loss [38]. This predisposition makes it difficult to study auditory aging, since even wild‐type controls show progressive deficits. Future studies will require refined behavioral and auditory phenotyping, older animals, and background strains without early‐onset hearing loss to better model age‐related interactions.

A further limitation of this study is the interpretation of Gene Ontology enrichment for poorly annotated genes. At least one third of the IMPC‐targeted genes have limited GO information, often based on computational or homology‐based inference rather than experimental evidence. This can bias enrichment toward broad functional categories and underrepresent less studied biological processes. For this reason, GO enrichment in this study should only be used to provide a general functional context.

4.6. A Path Forward

Previous work in models of early‐onset familial Alzheimer's disease has reported auditory dysfunction preceding cognitive symptoms. APP/PS1 and 5xFAD mice show elevated ABR thresholds and disrupted auditory processing at 2–3 months old, suggesting that hearing impairments may be among the earliest detectable changes [15, 16, 39]. However, results are not consistent. PS19 mice exhibit normal thresholds but altered latencies, pointing to central rather than peripheral changes [17]. These discrepancies highlight the complexity of auditory phenotypes in AD models and suggest that links between hearing and cognition may depend on genetic background or disease mechanism.

Our study moves beyond single transgenic models by using the IMPC dataset, which includes more than 9000 knockout lines assessed with standardized protocols. This provides a broader and less biased view of the genetic landscape linking hearing loss and cognitive‐emotional dysfunction. By identifying genes associated with both auditory and behavioral phenotypes, many not previously studied in this context, we offer new perspectives on contributors to cognitive impairment and dementia vulnerability. These findings also parallel human studies where hearing loss predicts not only cognitive decline [4, 5, 40] but also anxiety, depression, and social isolation [41, 42, 43, 44]. While most human work is observational, our results offer experimental support for the idea that ARHL and cognitive vulnerability may be mechanistically linked. At the same time, the diversity of effects suggests hearing loss is one of several interconnected factors in brain aging, not a singular cause.

The association between auditory dysfunction and behavioral abnormalities across many KO strains suggests that hearing loss may serve as a broader barometer of neural disruption. Identifying genes that link sensory and behavioral traits provides a foundation for new models of cognitive vulnerability and may enable earlier detection of those at risk for ADRD. Although our results reveal a consistent association between auditory and cognitive‐emotional abnormalities, confirming causal mechanisms will require additional research. One possibility is that both phenotypes arise from shared neuropathological pathways such as disrupted synaptic signaling, altered ion homeostasis, impaired cellular organization, or neuroinflammation. However, no single pathway emerged as a dominant driver. Instead, enrichment suggests a diffuse vulnerability spanning multiple aspects of cellular regulation and neuronal structure. This may indicate that auditory dysfunction acts less as a cause and more as a co‐occurring indicator of altered brain function, potentially shaped by indirect effects such as developmental compensation. Mechanistic studies using conditional knockouts, transcriptomic, and circuit‐level manipulations will be essential to clarify the pathways mediating the co‐occurrence of sensory and cognitive‐emotional abnormalities.

Projects like the IMPC are uniquely valuable for systematic discovery. Our findings also highlight the importance of expanding phenotyping pipelines to include refined auditory measures and cognitively demanding tasks tailored to species‐specific behavior. Such additions could better reveal links between sensory processing and brain function.

5. Conclusion

Together, these results emphasize auditory phenotyping as a window into cognitive dysfunction and provide a roadmap for future mechanistic studies. Identifying KO strains that link auditory and behavioral abnormalities lays the groundwork for targeted investigations of molecular and neural pathways underlying hearing loss and dementia. This approach may ultimately refine biomarker discovery and guide strategies for early intervention and risk stratification in aging‐related cognitive disorders.

Funding

This work was supported by the National Institutes of Health, R03AG081747.

David M. Rubenstein Fund for Hearing Research.

George T. Nager Professorship.

Disclosure

The data analyzed in this manuscript includes publicly available datasets from the IMPC and have not been previously published in the form presented here. The analyses described are original and are not under consideration elsewhere.

Ethics Statement

All animal experiments were conducted by IMPC partner institutions in accordance with international standards as well as their respective national and institutional guidelines for the care and use of animals.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Correlation between non‐auditory sensory phenotypes and top‐level mammalian phenotypic domains in KO mice. Bar plots show Spearman correlation coefficients between the presence of sensory abnormalities (A) vision/eye, or (B) vibrissae and top‐level phenotype categories defined by the Mammalian Phenotype Ontology. Each panel displays data from KO strains with abnormal vision/eye (1988 genes), or vibrissae (124 genes) phenotypes. Bars represent the strength of the association (Spearman's rho value), and asterisks (*) indicate statistically significant correlations after Benjamini‐Hochberg correction (p < 0.005).

Figure S2: Behavioral, cognitive, and emotional abnormalities are enriched among Vision/Eye genes. The figure shows the proportions of genes associated with abnormal general behavioral (A), emotional (B), or cognitive (C) phenotypes among KO lines with and without vision abnormalities. (D) shows odds ratios from Fisher's exact tests comparing the co‐occurrence of vision abnormalities with behavioral categories. Genes with vision abnormalities were more likely to exhibit behavioral abnormalities (OR = 1.86) and showed significantly increased odds of cognitive (OR = 1.71) and emotional (OR = 1.77) phenotypes. Asterisks indicate statistically significant results (p < 0.001).

Figure S3: Behavioral, cognitive, and emotional abnormalities are enriched among Vibrissa abnormalities. The figure shows the proportions of genes associated with abnormal general behavioral (A), emotional (B), or cognitive (C) phenotypes among KO lines with and without vibrissa abnormalities. (D) shows odds ratios from Fisher's exact tests comparing the co‐occurrence of vibrissa abnormalities with behavioral categories. Genes with vibrissa abnormalities were more likely to exhibit behavioral abnormalities (OR = 2.14) but did not have significantly elevated odds for cognitive (OR = 0.94) and emotional (OR = 1.06) abnormalities. Asterisks indicate statistically significant results (p < 0.001).

Data Availability Statement

The data that support the findings of this study are available in The International Mouse Phenotyping Consortium at https://www.mousephenotype.org, reference number RRID:SCR_006158. These data were derived from the following resources available in the public domain: The International Mouse Phenotyping Consortium, https://www.mousephenotype.org.