Sex differences in APOE- and PICALM-related cognitive profiles in healthy middle-aged adults

Introduction: The APOE $\epsilon 4$ and PICALM GG genotypes are strong genetic risk factors for Alzheimer’s disease. This study aimed to identify cognitive subgroups using unsupervised machine learning and to investigate the influence of APOE and PICALM genotypes on cognitive performance.

Material and methods: Cognitive, genetic and demographic data from 192 healthy middle-aged adults (50–63 years) from the PEARL-Neuro Database were analyzed using agglomerative hierarchical clustering. Neuropsychological tests included the California Verbal Learning Test, Raven’s Progressive Matrices, and the Edinburgh Handedness Inventory. Subsequent analyses used linear regression models to assess the effects of APOE, PICALM, and their interaction on cognitive outcomes.

Results: Two cognitive subgroups (better vs. worse performance) were identified for both females (n = 60/43) and males (n = 38/51). In women with lower cognitive performance, the presence of the APOE $\epsilon 3\epsilon 4$ allele was significantly associated with a higher number of perseverations (CVLT9: $p_{\text{FDR}}=0.02$, $R^2=0.18$) and lower recognition accuracy (CVLT12: $p_{\text{FDR}}=0.04$, $R^2=0.12$). A significant PICALM GG $\times$ education interaction was observed for fluid intelligence ($p_{\text{FDR}}=0.03$, $R^2=0.34$). In men with lower cognitive performance, the APOE $\epsilon 3\epsilon 4$ genotype was associated with lower fluid intelligence scores (RPM: $p_{\text{FDR}}=0.04$, $R^2=0.09$). Furthermore, significant APOE $\times$ PICALM interactions were found for verbal learning (CVLT1: $p_{\text{FDR}}=0.03$, $R^2=0.16$) as well as delayed cued recall (CVLT6: $p_{\text{FDR}}=0.03$, $R^2=0.12$; CVLT8: $p_{\text{FDR}}=0.03$, $R^2=0.13$).

Conclusion: This study revealed significant sex differences in gene–cognition interactions. In females with lower cognitive performance, the genotype APOE $\epsilon 3\epsilon 4$ was associated with poorer recognition, while the combined effects of APOE $\times$ PICALM in males were associated with weaker episodic memory. Although performance remained within normative ranges, these subtle differences may indicate early risk and warrant longitudinal monitoring.

1 Introduction

The APOE gene, which encodes apolipoprotein E, is a key genetic risk factor for Alzheimer’s disease (AD), with three common alleles: $\epsilon$2, $\epsilon$3, and $\epsilon$4 (Huang et al., 2019; Raulin et al., 2022). The variant $\epsilon$4, present in approximately 15% of the population, markedly increases the risk of AD compared to the genotype $\epsilon$3/$\epsilon$3 (Huang and Mahley, 2014). Carriers of one $\epsilon$4 allele have a three-fold higher risk of late-onset Alzheimer’s disease (LOAD), while homozygotes face up to a 15-fold increase (Hoogmartens et al., 2021; Liu et al., 2013). Although lifetime risk between sexes appears comparable from 55 to 85 years of age, women tend to show greater vulnerability earlier in life, suggesting age-dependent sex effects (Altmann et al., 2014; Neu et al., 2017; Eid et al., 2019). Previous studies suggest that the presence of the APOE $\epsilon$4 allele is associated with subtle cognitive impairments and an earlier onset of decline in younger populations, with episodic memory deficits potentially associated with hippocampal atrophy observed in healthy older adults and individuals with mild cognitive impairment (MCI) or AD (Lancaster et al., 2016; Sweet et al., 2012; Fan et al., 2019).

The PICALM (phosphatidylinositol binding clathrin assembly protein) gene has been identified as a risk locus for late-onset Alzheimer’s disease (LOAD) (Ando et al., 2022; Jansen et al., 2019). The single nucleotide polymorphism (SNP) rs3851179 affects the expression of PICALM and correlates with AD biomarkers, including cerebrospinal amyloid-$\beta$ and tau levels (Xu et al., 2020; Eid et al., 2019). Gene-wide analyses have also identified PICALM as a key locus associated with entorhinal cortical thickness, reinforcing its role in AD-related neurodegeneration (Furney et al., 2011). Furthermore, PICALM has been implicated in the timing and progression of cognitive decline, with the rs3851179 variant linked to an earlier onset of cognitive deterioration (Sweet et al., 2012). The A-allele is generally considered protective, while the G-allele confers an increased risk of AD (approximately 11%), potentially through altered gene expression and accelerated hippocampal atrophy (Ponomareva et al., 2020; Zeng et al., 2019; Xu et al., 2020). Carriers of the protective A-allele have shown reduced atrophy in the hippocampus, middle temporal gyrus, and the precuneus, as well as faster information processing and a lower risk of cognitive decline (Zeng et al., 2019). In individuals with MCI, the PICALM rs3851179 GG genotype has been associated with altered functional connectivity within the default mode network (Sun et al., 2017). Furthermore, rs3851179 has been reported to interact with the disease status to influence gray matter volume and cognition, with carriers of A-alleles exhibiting less putaminal atrophy and better global and executive functions, particularly in early-onset AD (Wu et al., 2022; Wu et al., 2024).

A synergistic adverse effect has also been observed between homozygosity for the risk allele (G) of PICALM rs3851179 and the presence of the APOE $\epsilon$4 allele, in individuals with early AD, resulting in reduced prefrontal volume and poorer performance on the Trail Making Test A, which is sensitive to processing speed (Morgen et al., 2014). Throughout, the association between PICALM and AD has been found predominantly in individuals carrying the risk allele APOE $\epsilon$4, supporting a synergistic interaction between the two genes in modulating the susceptibility of the disease (Jun, 2011). Combined risk genotypes have been associated with reduced episodic memory and a lower cerebral metabolic rate, particularly among carriers of APOE $\epsilon$4 (Barral et al., 2012; Chang et al., 2019).

The present study aimed to identify separate clusters of cognitive performance in males and females and to assess whether these groups differed in the genetic risk of AD. The APOE $\epsilon$4 allele has been associated with higher risk in women than in men, although the evidence in the studies remains mixed and not fully consistent (Sampedro et al., 2015; Sundermann et al., 2018). We further examined whether individual differences in neuropsychological performance could be explained by APOE $\epsilon$4 and PICALM GG genotypes, independently and interactively, and whether age and education moderated these associations. Identifying APOE–PICALM interactions can improve the early detection of individuals at risk of accelerated cognitive decline in late-onset AD (Bellenguez et al., 2022; Barral et al., 2012). Most of the existing evidence for sex-dependent APOE–PICALM interactions comes from clinical or pathology-enriched samples (MCI/AD). Direct investigations of sex-specific effects of PICALM on cognition in middle-aged cognitively healthy adults remain scarce. Although evidence on sex-dependent effects of PICALM remains limited, in clinical cohorts, the rs3851179 A-allele has been associated with better cognitive performance and a slower decline in older adults–particularly in men (Mengel-From et al., 2011). However, it remains unclear whether similar sex-dependent effects can be observed prior to the clinical manifestation of cognitive impairment. Therefore, examining the effects related to APOE–PICALM in cognitively healthy adults provides an opportunity to identify early sex-specific trajectories of cognitive aging.

The study utilized data from the publicly available PEARL-Neuro Database (Dzianok and Kublik, 2024). Previous analyses of this cohort did not find significant demographic or cognitive differences between carriers and non-carriers of APOE or PICALM risk alleles (Dzianok and Kublik, 2023a). Single APOE $\epsilon$4 carriers showed slower reaction times under high cognitive load, an effect mitigated by risk variants of PICALM, while EEG–fMRI data indicated reduced neural complexity and altered connectivity in at-risk individuals (Dzianok and Kublik, 2023b; Dzianok et al., 2024). To avoid redundancy, the variables that overlapped from these previous analyses were excluded from the present study. Previous Polish studies on APOE focused on its association with LOAD, with no additional effects from other loci (Styczynska et al., 2003; Religa et al., 2003).

To uncover latent patterns, we reanalyzed neuropsychological data using unsupervised machine learning, which identifies the structure in unlabeled data without predefined categories (Dalmaijer et al., 2022). Such data-driven approaches–including clustering and dimensionality reduction (e.g., principal component analysis)–are increasingly applied in behavioral genetics and neuropsychology to reveal non-linear associations between cognitive phenotypes and genetic variation (Wang et al., 2024; Scheltens et al., 2016).

2 Materials and methods

2.1 Dataset and participants

Study participants were selected from the PEARL-Neuro Database. A detailed description of the trial has been described elsewhere (Dzianok and Kublik, 2024). The database contains genetic information on APOE and PICALM genes. APOE (rs429358/rs7412, necessary to identify the main isoforms $\epsilon$2, $\epsilon$3 and $\epsilon$4) and PICALM (rs3851179) alleles were genotyped using the traditional Sanger sequencing method, a reliable and well-established DNA sequencing approach (Dzianok and Kublik, 2024). In addition, this database includes psychometric data, basic demographic information, and health data in a group of 192 healthy middle-aged individuals (aged 50–63 years). Of the participants, 77.6% (n = 149) reported higher education, 10.42% (n = 20) had secondary education, and 2.60% (n = 5) had partial higher education, with data missing for 18 individuals. Table 1 presents descriptive statistics for all participants and stratified by sex, while Table 2 summarizes the distribution of APOE and PICALM genotypes in the male and female groups. These data represent the characteristics of the sample prior to further preprocessing. Only variables assessing cognitive functioning were selected from the dataset, including the California Verbal Learning Test (CVLT), Raven’s Progressive Matrices (RPM), and the Edinburgh Handedness Inventory (EHI), which served as input features for the unsupervised clustering. The definitions of the selected parameters are provided in Table 3.

Table 1

Table 1. Descriptive statistics for all participants and stratified by sex.

Table 2

Table 2. Distribution of APOE and PICALM genotypes by sex.

Table 3

Table 3. Selected cognitive function variables from the dataset.

A description of the methods is given below:

• CVLT - is a widely used tool in clinical and research settings to assess verbal learning and memory. It enables the evaluation of various cognitive processes, including learning between repetitions, serial position effects, semantic clustering, intrusions, and proactive interference. The test was designed with ecological validity, incorporating tasks that reflected daily activities, such as recalling shopping lists (Delis et al., 2008; Elwood, 1995; Łojek et al., 2010).

• RPM - is a nonverbal test designed to assess general cognitive ability, particularly abstract reasoning and problem-solving skills. The research findings indicate that RPM can be considered a relatively independent test of cultural factors to measure fluid intelligence (Raven et al., 2003; Raven, 2008; Jaworowska and Szustrowa, 2010). The allotted time to complete the test was set at 30 min, replacing the unlimited time frame used in the original version (Dzianok and Kublik, 2024).

• EHI - is a brief tool designed to assess handedness on a quantitative scale. This distinction is relevant due to its association with individual differences in neuropsychological and functional brain characteristics (Edlin et al., 2015).

2.2 Machine learning approach and statistical analysis

The diagram below presents an overview of the analytical workflow used in this study (see Figure 1).

• Step 1: The dataset was comprehensively pre-processed statistically. In the initial step, z-score standardization was applied to normalize the scale of the variables and ensure their comparability. Each feature was transformed to have a mean of zero and a standard deviation of one, which was essential due to the use of the Euclidean distance metric in subsequent analyses.

• Step 2: Principal Component Analysis (PCA) was used to reduce dimensionality and identify the main sources of variance within the dataset (Greenacre et al., 2022). The first two principal components were used to illustrate the clustering results.

• Step 3: Agglomerative hierarchical clustering was applied to explore the latent structure within the neuropsychological dataset. This unsupervised learning method is well suited to small and moderately sized datasets typical of clinical research, where group labels are often unknown or hypothetical (Li et al., 2022). Clustering was conducted using the AgglomerativeClustering algorithm from the scikit-learn library with Euclidean distance and Ward’s linkage, which minimizes within-cluster variance and is appropriate for continuous cognitive data. To compute the distance between observations, the standard Euclidean distance was used, defined as:

$d\left(\mathbf{x},\mathbf{y}\right)=\sqrt{{\displaystyle \sum _{i=1}^n}{\left(x_i-y_i\right)}^2}$

where $\mathbf{x}$ and $\mathbf{y}$ are two observations in $n$-dimensional space.

Figure 1

Figure 1. Overview of the machine learning and statistical analysis workflow.

Agglomerative clustering has been shown to be useful in genotype-to-phenotype studies (Sasirekha and Baby, 2013). Although this analysis originally aimed to cluster based on genotype, the limited number of rare variants (e.g., individuals carrying two $\epsilon 4$ alleles) precluded meaningful comparisons. Therefore, a reverse approach was adopted–clustering based on cognitive performance (phenotype) and subsequently examining the distribution of APOE and PICALM genotypes within clusters. The ambiguous genotype APOE $\epsilon 2\epsilon 4$ was excluded from further analyses. The number of clusters was estimated based on dendrogram inspection and by maximizing the silhouette score, a metric that evaluates how well each observation fits within its assigned cluster compared to others. The hierarchical dendrogram was cut at a distance level of approximately 26.161 for males and 29.904 for females, yielding two clusters determined by the parameter n_clusters = 2 in the Agglomerative Clustering model. Hierarchical clustering was performed separately for males and females to account for established sex-specific patterns in cognitive performance and APOE/PICALM effects, thus avoiding potential bias or dilution of within-sex structure that might arise if sex were treated only as a covariate. To assess whether the sample sizes obtained through clustering were adequate, an a priori power analysis was conducted. Assuming $\alpha =0.05$, power = 0.80, and medium-to-large effect size (Cohen’s d = 0.6), the estimated minimum sample size per group was approximately 45 participants, totaling 89 for balanced groups (Brydges, 2019; Lakens, 2022). In our study, the resulting clusters comprised 43 and 60 women, and 38 and 51 men, that align with methodological recommendations for unsupervised learning approaches, which suggest a minimum of 20–30 participants per subgroup (Dalmaijer et al., 2022).

• Step 4: To assess differences between clusters, cognitive variables were compared using parametric or nonparametric tests, depending on data distribution (Shapiro–Wilk) and variance homogeneity (Levene’s test). Student’s or Welch’s t-tests were applied for normally distributed data, and the Mann–Whitney U test for non-normal distributions. Differences in genotype frequencies were examined using chi-square or Fisher’s exact tests. Mean neuropsychological scores across the four clusters (two female, two male) were compared using ANOVA with Tukey’s post hoc tests to identify significant differences between-group.

• Step 5 Linear regression analyses were conducted within each cluster to assess the effects of genetic risk genotypes–APOE $(\epsilon 3\epsilon 4)$ and PICALM GG–on neuropsychological performance. Three main models for each cognitive variable were estimated: including APOE, PICALM, and their interaction (APOE $\times$ PICALM). Additional models examined interactions with age and education, including higher-order terms (e.g., APOE $\times$ age $\times$ education). For each model, regression coefficients $(\beta )$, p-values and coefficients of determination $(R^2)$ were recorded. To account for multiple comparisons, all p-values obtained from the regression analyses were adjusted using the False Discovery Rate (FDR) correction according to the Benjamini–Hochberg procedure (Benjamini and Hochberg, 1995). The FDR-adjusted p-values $(p_{\mathit{FDR}})$ were calculated separately for the better- and worse-performing cognitive clusters. Only effects with $p_{\mathit{FDR}}<0.05$ were considered statistically significant. The analyses were performed separately for the cognitively “better” and “worse” clusters, which reflect relative cognitive profiles rather than clinical categories. Clustering was performed in the entire neuropsychological dataset to preserve the structure of the data-driven group, while regression analyses were limited to carriers of single risk alleles (APOE $\epsilon$3/$\epsilon$4 and PICALM GG) to ensure consistency between sexes. All analyses were performed in Python 3.0; the code is available on request.

3 Results

3.1 Clustering analysis for female and male participants

The cluster analysis conducted separately for female and male participants revealed two distinct groups within each sex. Visualization using principal component analysis (PCA) showed clear separation between clusters in both females and males (Figure 2). Hierarchical clustering further supported the presence of two main clusters within each sex (Figure 3). No significant differences were observed between clusters in age or education (Table 4). As expected, since the clustering was based on measures of neuropsychological performance, all neuropsychological variables differed significantly between the clusters $(p<0.05)$.Therefore, these findings serve primarily to characterize the cognitive profiles of the newly identified clusters in both female and male. In both groups, the participants in Cluster 1 consistently outperformed those in Cluster 0 in all CVLT measures. This included higher total learning scores across trials (CVLT_1), better performance in both initial and final learning trials (CVLT_2–CVLT_3), and superior results in delayed recall conditions–covering short-delay free recall and long-delay free and cued recall tasks (CVLT_5–CVLT_8). Individuals in Cluster 1 also made fewer errors overall, including perseveration errors (CVLT_9) and intrusion errors in both free (CVLT_10) and cued recall (CVLT_11). Recognition performance also favored Cluster 1, reflected in a higher number of correct hits (CVLT_12) and fewer false alarms (CVLT_13). Beyond verbal memory, participants in Cluster 1 also demonstrated higher fluid intelligence scores (RPM) and a higher degree of right-handedness as measured by EHI.

Figure 2

Figure 2. Clustering based on principal component analysis (PCA) in female (left panel) and male (right panel) groups, illustrating group separation in a two-dimensional space.

Figure 3

Figure 3. Hierarchical clustering dendrograms obtained using Ward’s method for the female (upper panel) and male (lower panel) groups. Each data point starts as an individual cluster, and clusters are progressively merged based on similarity.

Table 4

Table 4. Characteristics of participants and genotype distribution by cognitive cluster in female and male groups together.

In all cognitive measures, one-way ANOVA tests revealed highly significant group effects (all p < 0.001), with post hoc Tukey comparisons indicating that the “female better” cluster consistently outperformed the other groups on nearly all CVLT variables, similar patterns were observed for the RPM and EHI measures, although the differences between the higher-performing clusters were not statistically significant (see Supplementary Materia1 1.2) Additional boxplots and histograms illustrating the distributions of individual neuropsychological measures (CVLT_1–CVLT_13, RPM, and EHI) for each subgroup are provided in the Supplementary Material 1.3 and 1.5 to facilitate the evaluation of within-group variability and possible ceiling or floor effects.

Fisher’s exact test revealed no statistically significant differences in genotype distribution in clusters in either sex (APOE: $p=0.68$ for females, $p=0.41$ for males; PICALM: $p=0.53$ for females, $p=0.76$ for males). This indicates that cluster membership was not significantly associated with the APOE or PICALM genotypes in the present sample. Detailed results are presented in Table 4 and illustrated in Figure 4 (upper and lower panels for females and males, respectively).

Figure 4

Figure 4. Distribution of APOE and PICALM genotypes by cluster for the female (upper panel) and male (lower panel) groups. Genotype counts are shown separately for the cognitive “better” and “worse” clusters within each sex.

To account for potential somatic influences on cognitive performance, additional analyses of basic clinical indicators were conducted; however, blood count data were available for only a subset of participants, substantially limiting interpretability (see Supplementary Material 1.4).

3.2 Genetic influence of APOE and PICALM and their interaction on cognitive performance - in female’s group

3.2.1 Worse-performing cognitive cluster

Only the results that survived the Benjamini–Hochberg false discovery rate (FDR) correction $(p_{\text{FDR}}<0.05)$ are reported in this section; non-significant associations ($p<0.05$ but $p_{\text{FDR}}>0.05$) were omitted for clarity. The percentages reported in parentheses refer to the proportion of explained variance $(R^2)$ for each model, as shown in Table 5. All other interactions are reported in the Supplementary Material 1.1.

Table 5

Table 5. Regression models for APOE and PICALM effects in females: comparison between better (blue) and worse (red) cognitive clusters.

The presence of the APOE $\epsilon 3\epsilon 4$ allele was associated with a higher number of perseverations (CVLT_9) and lower recognition performance (CVLT_12), with the genetic factor explaining approximately 18% and 13% of the variance in neuropsychological test performance, respectively. Significant APOE $\times$ age interactions were observed for recognition performance (CVLT_12), and an APOE $\times$ education interaction remained significant for false recognition (CVLT_13), with the respective models explaining approximately 28% and 23% of the variance (see Table 5). The PICALM GG genotype was significantly associated with a lower number of intrusion errors (CVLT_10; $R^2=0.23$, CVLT_11; $R^2=0.16$). Additionally, a significant PICALM $\times$ education interaction was observed for fluid intelligence (RPM), which explained approximately 35% of the variance. Finally, a three-way interaction of APOE $\times$ PICALM $\times$ age was detected for long-delay cued recall (CVLT_8), explaining about 22% of variance. Furthermore, a significant four-way interaction of APOE $\times$ PICALM $\times$ age $\times$ education was detected for CVLT_5 (short-delay free recall; 50% of variance explained) and CVLT_8 (long-delay cued recall; 51% of variance explained) (see Table 5).

3.2.2 Better-performing cognitive cluster

Several nominally significant associations were initially observed in the better-performing cognitive cluster, including interactions involving APOE, PICALM, age, and education. However, after applying the FDR (Benjamini–Hochberg) correction for multiple comparisons, none of these effects remained statistically significant ($p_{\mathit{FDR}}$ > 0.05). Data are presented in Table 5.

3.3 Genetic influence of APOE and PICALM and their interaction on cognitive performance - in male’s group

3.3.1 Worse-performing cognitive cluster

In this cluster, the APOE $\epsilon$4 allele was negatively associated with fluid intelligence (RPM), explaining approximately 10% of the variance, whereas the interaction with education increased the explained variance to around 16% (see Table 6). In the three-way APOE $\times$ age $\times$ education analysis, significant effects were observed for several CVLT measures related to short- and long-delay recall (CVLT_5–CVLT_8), with models that explained between 27% and 36% of variance. For the PICALM gene, no significant main effects were observed; however, interactions with education accounted for approximately 13% of the variance in false recognition errors (CVLT_13) and perseveration errors (CVLT_9), respectively. Significant APOE $\times$ PICALM interactions were observed in several memory measures (CVLT_1, CVLT_3, CVLT_6, CVLT_8, CVLT_13), with models explaining between 13% and 23% of the variance (see Table 6). Furthermore, significant three-way APOE $\times$ PICALM $\times$ age interactions were found for CVLT_11 and CVLT_13, with explained variance ranging from 52% to 58%. Similarly, APOE $\times$ PICALM $\times$ education interactions were observed for CVLT_5–CVLT_8, representing 25%–32% of the variance in recall performance. Finally, a significant four-way interaction APOE $\times$ PICALM $\times$ age $\times$ education was found for CVLT_11 (intrusion errors), which explained approximately 57% of the variance (see Table 6).

Table 6

Table 6. Regression models for APOE and PICALM effects in males: comparison between better (blue) and worse (red) cognitive clusters.

3.3.2 Better-performing cognitive cluster

In this cluster, a significant three-way interaction (PICALM GG $\times$ age $\times$ education) was observed only for handedness $(R^2=0.43)$. Although several nominal associations were observed before correction (e.g., CVLT_4, CVLT_5), none remained a significant adjustment of the FDR (all $p_{\mathit{FDR}}$> 0.05). Data are presented in Table 6.

4 Discussion

4.1 Cognitive profiles and clustering patterns in female group’s

Within the lower-performing female cluster, carriers of the genotype APOE$\epsilon 3\epsilon$4 showed significantly lower recognition precision and a higher number of perseverative errors. No such associations were observed in the higher-performing group. The APOE $\epsilon$4 allele has been consistently associated with recognition difficulties in both clinical and healthy groups, including a faster decline in word recognition and overall poorer recognition performance (Hirono et al., 2003; Haley et al., 2010). Perseverations, commonly observed in AD and list-learning tasks, are believed to reflect deficits in working memory and response monitoring (Pakhomov et al., 2018; Miozzo et al., 2013), consistent with reports of reduced working memory in cognitively healthy carriers $\epsilon$4, particularly under interference-control demands (Reinvang et al., 2010). However, given that the participants were cognitively intact, a certain level of perseverations remains within normative limits (Łojek et al., 2010), especially with extended word lists. Thus, these findings should be interpreted with caution. Although the PICALM GG genotype is considered a risk variant, our findings–showing fewer intrusion errors among GG carriers–may reflect delayed expression of its adverse effects or the influence of cognitive reserve (Pinto and Tandel, 2016). In line with previous reports indicating that cognitive reserve can mitigate the detrimental impact of the APOE $\epsilon$4 allele on memory performance (Pettigrew et al., 2013), a similar compensatory mechanism may also operate in the case of the PICALM GG genotype. Alternatively, a form of antagonistic pleiotropy, similar to that proposed for the APOE $\epsilon$4 allele (Fan et al., 2019; Tuminello and Han, 2011; Rusted et al., 2013), cannot be excluded and warrant further investigation.

4.2 Cognitive profiles and clustering patterns in male group’s

The presence of the APOE $\epsilon$3/$\epsilon$4 genotype was associated with lower performance on the RPM task, reflecting reduced fluid intelligence. Our findings indicate that the APOE $\times$ PICALM interaction significantly affects multiple stages of verbal learning and episodic memory, suggesting that PICALM may modulate APOE-related cognitive vulnerability (Morgen et al., 2014), and in the lower-performing male cluster this pattern resembles a “flip-flop effect,” where the direction of genetic influence depends on the allelic context of another gene (Engelman et al., 2013; Liu et al., 2023; Fan et al., 2019). However, since cognitively healthy male carriers $\epsilon$4 also show poorer episodic memory (Sundermann et al., 2018), it is not clear whether the observed deficits reflect the combined effect of APOE $\times$ PICALM or APOE $\epsilon$4 alone. Consistent with previous work, carriers of APOE $\epsilon$4 exhibit reduced performance in episodic memory, executive functions, and nonverbal cognition, and APOE interacts with early-life cognitive ability to influence processing speed and variability of reaction time in older adults (Schultz et al., 2008; Liu et al., 2013; Izaks et al., 2011; Luciano et al., 2009). Longitudinal studies are needed to clarify these relationships.

In the lower-performing male cluster, carriers of the APOE $\epsilon$4 and PICALM GG genotypes showed increased intrusion errors (CVLT_11) and false recognitions (CVLT_13), but only in interaction models including age and education (APOE $\times$ PICALM $\times$ age; APOE $\times$ PICALM $\times$ age $\times$ education). Older age combined with these risk genotypes was associated with increased intrusions and false recognitions, reflecting a greater susceptibility to memory interference and reduced retrieval accuracy. This pattern may reflect subtle deficits in inhibitory control or memory monitoring, consistent with prior evidence linking intrusion errors in verbal memory tasks to an increased risk of progression from normal cognition to MCI and early dementia (Thomas et al., 2018; Torres et al., 2019). Age can modulate the cognitive impact of the APOE $\epsilon$4 allele, with younger carriers showing relatively preserved recall and older adults showing decline (Fan et al., 2019; Caselli et al., 2009; Jochemsen et al., 2012), although results remain mixed (Bunce et al., 2014).

4.3 Sex-specific genetic interactions and cognitive performance

Analysis of APOE-PICALM interactions revealed different sex-dependent patterns, with males with cognitively lower performance showing deficits related to APOE $\epsilon$4 in fluid intelligence, and combined effects of APOE-PICALM on episodic memory. In females, the associations mainly involved APOE-related influences on recognition and executive-control measures, including increased perseverative errors among $\epsilon$4 carriers. Although our study examined young, cognitively healthy adults, the sex-dependent APOE–PICALM patterns may represent early, subclinical manifestations of cognitive vulnerabilities that only later become clinically apparent. These findings are consistent with previous evidence that the detrimental effect of the APOE $\epsilon$4 allele on verbal memory is primarily evident in cognitively normal men, while women tend to maintain better memory performance despite a comparable amyloid burden (Sundermann et al., 2018). This pattern may reflect enhanced compensatory mechanisms in women, which could delay the manifestation of cognitive deficits related to APOE $\epsilon$4 until more advanced stages of the disease (Sundermann et al., 2018; Jack Jr et al., 2013). In general, males more often exhibited higher-order effects (e.g., gene $\times$ age $\times$ education) with greater explanatory power, suggesting that cognitive performance in men may depend on a more complex gene–environment interplay, but such sex-dependent interaction patterns may help explain part of the missing heritability and variability in cognitive aging trajectories (Singhal et al., 2023; Altmann et al., 2014; Neu et al., 2017; Sundermann et al., 2018). In both lower-performing clusters, education emerged as a key modifier of genetic effects. Among females, a significant PICALM_GG $\times$ education interaction was found for fluid intelligence (RPM), suggesting that higher educational attainment mitigated the negative impact of the PICALM GG risk genotype, possibly reflecting a compensatory role of cognitive reserve (Pinto and Tandel, 2016). In males, the APOE $\epsilon$4 allele was negatively associated with fluid intelligence, with education increasing the explained variance from approximately 10%–16%. Furthermore, APOE $\times$ PICALM $\times$ education interaction was observed for short- and long-term recall (CVLT_5–8), indicating that higher education buffered the adverse cognitive effects of genetic risk. This pattern aligns with evidence that education and related aspects of cognitive reserve can moderate the impact of genetic vulnerability on memory and overall cognitive functioning in older adults (Hsu et al., 2025; Walsemann et al., 2025; Pettigrew et al., 2023; Eid et al., 2019). Similar gene–education interactions have also been reported in AD cohorts, where the cognitive and neural effects of these loci are influenced by age, disease stage, and cognitive reserve (Morgen et al., 2014; Chang et al., 2021; Vonk et al., 2019; Liu et al., 2023; Fan et al., 2019). The relatively high educational level of the current sample may therefore have reduced between-group variability. With greater diversity in educational attainment, cognitive score distributions might have differed, as genetic factors explain more variance in individuals with lower education. Previous research shows that higher education and other cognitive-reserve proxies support better cognitive performance, helping to maintain functioning despite aging or pathology (Le Carret et al., 2003; Panico et al., 2023). It should be noted that educational attainment, although often treated as environmental, is partly heritable (40%–60%) and a recognized modifiable risk factor for dementia (Ayorech et al., 2017; Livingston et al., 2024). The main effects of handedness were not found in worse-performing clusters, consistent with previous large-scale studies that did not show a relationship between handedness and the genetic risk of AD (Hubacek et al., 2013; Abuduaini et al., 2024). Furthermore, the observed associations between genetic variants and fluid intelligence (RPM) highlight their value as an early marker of gene–cognition interactions. Although less commonly applied in MCI or AD diagnostics (Ruchinskas, 2019; Boccardi et al., 2022), intelligence measures capture interindividual variability in aging and are strongly heritable, with broad and largely domain-general genetic influences, since most genetic effects on cognition are considered general rather than domain-specific (Plomin and Deary, 2015; Plomin and Von Stumm, 2018). Finally, our findings support the decision to analyze males and females separately. Sex-specific pathways in cognitive aging–linked to inflammatory, metabolic and microglial differences–may contribute to increased vulnerability of women to neurodegeneration (Hanamsagar and Bilbo, 2016; Holland et al., 2013; Arnold et al., 2020) and modify the risk of AD-related cognitive decline (Buckley et al., 2018). Some of our findings contrast with previous reports suggesting a stronger effect of APOE $\epsilon$4 in women than in men on cognitive outcomes in aging and AD, a discrepancy that may reflect differences in sample characteristics, race, diagnostic criteria, study design, and the specific cognitive domains assessed (Altmann et al., 2014; Neu et al., 2017; Barnes et al., 2013; Sundermann et al., 2018; Eid et al., 2019; Fan et al., 2019).

Several limitations should be considered when interpreting the present findings, particularly given the exploratory nature of this study and its focus on genetic influences on cognitive functioning. First, the relatively small sample size, particularly after stratification by sex and cognitive cluster, limited the statistical power to detect subtle effects and restricted the generalizability of the findings. However, an a priori power analysis indicated that the resulting cluster sizes were close to recommended thresholds, exceeding the minimum commonly cited for subgroup analyses. Thus, while the sample was not optimal for detecting smaller effects, it was adequate for exploratory regression analyses at the cluster level. Replication in larger datasets is essential. Furthermore, the scope of neuropsychological evaluation was limited, focusing primarily on selected cognitive domains such as memory, and biomarker data were not included. Second, the phenotype-to-genotype approach may have introduced a selection bias. Although this strategy helps identify specific cognitive effects of known variants, it may overlook subtler phenotypic variation and lead to biased estimates of variant penetrance and pathogenicity (Wilczewski et al., 2023). Third, this study used only the Agglomerative Clustering algorithm for unsupervised learning. Alternative approaches, including Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and K-means, were tested but produced suboptimal results (see Supplementary Material 1.6). DBSCAN was highly sensitive to parameter settings and did not separate clusters with heterogeneous densities, often collapsing the data into a single dominant group (Ester et al., 1996). The K-means, which assume spherical clusters and equal variance, yielded unstable solutions and small clusters below the power analysis threshold, limiting the interpretability (Chong, 2021; Dalmaijer et al., 2022; Brydges, 2019). Future research should include a broader spectrum of cognitive and non-cognitive variables, such as personality traits (e.g., neuroticism), to better capture gene–behavior interactions (Hindley et al., 2023). Although no association has been observed between the APOE $\epsilon$4 allele and affective symptoms in Polish patients with AD (Gabryelewicz et al., 2002), incorporating such measures may improve understanding of cognitive–emotional profiles in preclinical stages. Moreover, explainable machine learning methods (e.g., SHapley Additive exPlanations values) could help unravel the relative impact of APOE $\epsilon$4, PICALM GG, and their interplay with age, education and other modifiable risk factors (Lundberg and Lee, 2017; Fan et al., 2019).

5 Conclusion

This study suggests that the effects of the APOE $\epsilon 3\epsilon$4 allele and the PICALM GG variant on cognitive functioning may vary depending on sex. In lower-performing females, the APOE $\epsilon 3\epsilon$4 allele was associated with less precise recognition and more perseverative errors; in males, with reduced fluid intelligence. Additionally, in males, APOE $\times$ PICALM interactions affected delayed recall and recognition, indicating combined genetic effects on memory retrieval. A higher number of significant interactions for the APOE variant was observed in both the female and male groups. The PICALM GG genotype was less frequently involved in significant interactions independently, but its relevance increased significantly when interacting with APOE, highlighting the utility of such interaction-based analyses in future AD research. In particular, three-way interactions (gene $\times$ age $\times$ education) were observed more frequently in the male group. The clustering approach revealed subtle cognitive differences that remain within normative ranges. Longitudinal studies are needed to determine whether these patterns reflect early vulnerability or normal variability. Genetic effects should not be viewed as directly “encoding” cognition; instead, they influence cognitive outcomes through complex molecular, cellular, and environmental interactions.

Data availability statement

Publicly available datasets were analyzed in this study. This data can be found here: https://openneuro.org/datasets/ds004796/versions/1.0.4. The code is available from the authors upon request.

Ethics statement

The dataset used in the present study was previously approved by the Bioethics Committee of the Nicolaus Copernicus University in Toruń at the Ludwik Rydygier Collegium Medicum in Bydgoszcz, Poland (approval number: KB 684/2019). All participants (N = 200) provided written informed consent and signed an extended study information form, which included detailed information on data privacy, pseudonymization, and anonymization procedures applied for the purposes of analyses and publications related to this research project. Of the 200 participants who participated in the study, 192 signed an addendum, agreeing to make the research data publicly available in the open scientific database. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Author contributions

AB: Conceptualization, Formal Analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review and editing. PT: Formal Analysis, Software, Writing – review and editing. MH: Methodology, Writing – review and editing. DR: Supervision, Writing – review and editing.

Funding

The author(s) declared that financial support was not received for this work and/or its publication.

Acknowledgements

We express our sincere gratitude to Patrycja Dzianok and Ewa Kublik from the Nencki Institute of Experimental Biology for creating and sharing the Polish Electroencephalography, Alzheimer’s Risk-genes, Lifestyle and Neuroimaging (PEARL-Neuro) Database. We especially acknowledge the Nencki Institute for making this dataset available as part of the open-data movement, a global initiative aimed at promoting the sharing of data and information in a transparent and accessible manner.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fragi.2025.1694701/full#supplementary-material

References

Abuduaini, Y., Chen, W., and Kong, X.-Z. (2024). Handedness in alzheimer’s disease: a systematic review. Brain Res. 1840, 149131. doi:10.1016/j.brainres.2024.149131

PubMed Abstract | CrossRef Full Text | Google Scholar

Altmann, A., Tian, L., Henderson, V. W., Greicius, M. D., and Investigators, A. D. N. I. (2014). Sex modifies the apoe-related risk of developing alzheimer disease. Ann. Neurol. 75, 563–573. doi:10.1002/ana.24135

PubMed Abstract | CrossRef Full Text | Google Scholar

Ando, K., Nagaraj, S., Küçükali, F., De Fisenne, M.-A., Kosa, A.-C., Doeraene, E., et al. (2022). Picalm and Alzheimer’s disease: an update and perspectives. Cells 11, 3994. doi:10.3390/cells11243994

PubMed Abstract | CrossRef Full Text | Google Scholar

Arnold, M., Nho, K., Kueider-Paisley, A., Massaro, T., Huynh, K., Brauner, B., et al. (2020). Sex and apoe ɛ4 genotype modify the alzheimer’s disease serum metabolome. Nat. Commun. 11, 1148. doi:10.1038/s41467-020-14959-w

PubMed Abstract | CrossRef Full Text | Google Scholar

Ayorech, Z., Krapohl, E., Plomin, R., and Von Stumm, S. (2017). Genetic influence on intergenerational educational attainment. Psychol. Sci. 28, 1302–1310. doi:10.1177/0956797617707270

PubMed Abstract | CrossRef Full Text | Google Scholar

Barnes, L., Arvanitakis, Z., Yu, L., Kelly, J., De Jager, P., and Bennett, D. (2013). Apolipoprotein e and change in episodic memory in blacks and whites. Neuroepidemiology 40, 211–219. doi:10.1159/000342778

PubMed Abstract | CrossRef Full Text | Google Scholar

Barral, S., Bird, T., Goate, A., Farlow, M., Diaz-Arrastia, R., Bennett, D., et al. (2012). Genotype patterns at picalm, cr1, bin1, clu, and apoe genes are associated with episodic memory. Neurology 78, 1464–1471. doi:10.1212/WNL.0b013e3182553c48

PubMed Abstract | CrossRef Full Text | Google Scholar

Bellenguez, C., Küçükali, F., Jansen, I. E., Kleineidam, L., Moreno-Grau, S., Amin, N., et al. (2022). New insights into the genetic etiology of Alzheimer’s disease and related dementias. Nat. Genetics 54, 412–436. doi:10.1038/s41588-022-01024-z

PubMed Abstract | CrossRef Full Text | Google Scholar

Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Statistical Soc. Series B Methodol. 57, 289–300. doi:10.1111/j.2517-6161.1995.tb02031.x

CrossRef Full Text | Google Scholar

Boccardi, M., Monsch, A. U., Ferrari, C., Altomare, D., Berres, M., Bos, I., et al. (2022). Harmonizing neuropsychological assessment for mild neurocognitive disorders in Europe. Alzheimer’s Dementia 18, 29–42. doi:10.1002/alz.12365

PubMed Abstract | CrossRef Full Text | Google Scholar

Brydges, C. R. (2019). Effect size guidelines, sample size calculations, and statistical power in gerontology. Innovation Aging 3, igz036. doi:10.1093/geroni/igz036

PubMed Abstract | CrossRef Full Text | Google Scholar

Buckley, R. F., Mormino, E. C., Amariglio, R. E., Properzi, M. J., Rabin, J. S., Lim, Y. Y., et al. (2018). Sex, amyloid, and apoe ɛ4 and risk of cognitive decline in preclinical alzheimer’s disease: findings from three well-characterized cohorts. Alzheimer’s Dementia 14, 1193–1203. doi:10.1016/j.jalz.2018.04.010

PubMed Abstract | CrossRef Full Text | Google Scholar

Bunce, D., Bielak, A. A., Anstey, K. J., Cherbuin, N., Batterham, P. J., and Easteal, S. (2014). Apoe genotype and cognitive change in young, middle-aged, and older adults living in the community. J. Gerontol. Ser. A Biomed. Sci. Med. Sci. 69, 379–386. doi:10.1093/gerona/glt103

PubMed Abstract | CrossRef Full Text | Google Scholar

Caselli, R. J., Dueck, A. C., Osborne, D., Sabbagh, M. N., Connor, D. J., Ahern, G. L., et al. (2009). Longitudinal growth modeling of cognitive aging and the apoe e4 effect. N. Engl. J. Med. 361, 255–263. doi:10.1056/NEJMoa0809437

PubMed Abstract | CrossRef Full Text | Google Scholar

Chang, Y.-T., Huang, C.-W., Huang, S.-H., Hsu, S.-W., Chang, W.-N., Lee, J.-J., et al. (2019). Genetic interaction is associated with lower metabolic connectivity and memory impairment in clinically mild Alzheimer’s disease. Genes Brain Behav. 18, e12490. doi:10.1111/gbb.12490

PubMed Abstract | CrossRef Full Text | Google Scholar

Chang, Y.-L., Zhuo, Y.-Y., and Luo, D.-H. (2021). Education moderates the negative effect of apolipoprotein E ɛ4 on response inhibition in older adults. J. Alzheimer’s Dis. 82, 1147–1157. doi:10.3233/JAD-210183

PubMed Abstract | CrossRef Full Text | Google Scholar

Chong, B. (2021). K-means clustering algorithm: a brief review. Acad. J. Comput. Inf. Sci. 4, 37–40.

Google Scholar

Dalmaijer, E. S., Nord, C. L., and Astle, D. E. (2022). Statistical power for cluster analysis. BMC Bioinform. 23, 205. doi:10.1186/s12859-022-04675-1

PubMed Abstract | CrossRef Full Text | Google Scholar

Delis, D. C., Kaplan, E., Kramer, J. H., and Ober, B. A. (2008). California verbal learning test (cvlt) (ECPA United States).

Google Scholar

Dzianok, P., and Kublik, E. (2023a). Altered granulocyte count and erythrocyte measures in middle-aged, healthy carriers of apoe and picalm risk genes for Alzheimer’s disease. Acta Neurobiol. Exp. 83, 127–139. doi:10.55782/ane-2023-012

PubMed Abstract | CrossRef Full Text | Google Scholar

Dzianok, P., and Kublik, E. (2023b). Alzheimer’s disease risk alleles of picalm gene may accelerate cognitive (memory and attention) performance in middle-aged healthy apoe-ɛ4 carriers. Alzheimer’s Dementia 19, e067545. doi:10.1002/alz.067545

CrossRef Full Text | Google Scholar

Dzianok, P., and Kublik, E. (2024). Pearl-neuro database: EEG, fMRI, health and lifestyle data of middle-aged people at risk of dementia. Sci. Data 11, (1) 276. doi:10.1038/s41597-024-03106-5

PubMed Abstract | CrossRef Full Text | Google Scholar

Dzianok, P., Wojciechowski, J., Wolak, T., and Kublik, E. (2024). Alzheimer’s disease-like features in resting state eeg/fmri of cognitively intact and healthy middle-aged apoe/picalm risk carriers. bioRxiv. doi:10.1101/2024.06.20.599857

CrossRef Full Text | Google Scholar

Edlin, J. M., Leppanen, M. L., Fain, R. J., Hackländer, R. P., Hanaver-Torrez, S. D., and Lyle, K. B. (2015). On the use (and misuse?) of the edinburgh handedness inventory. Brain Cognition 94, 44–51. doi:10.1016/j.bandc.2015.01.003

PubMed Abstract | CrossRef Full Text | Google Scholar

Eid, A., Mhatre, I., and Richardson, J. R. (2019). Gene-environment interactions in Alzheimer’s disease: a potential path to precision medicine. Pharmacol. Therapeutics 199, 173–187. doi:10.1016/j.pharmthera.2019.03.005

PubMed Abstract | CrossRef Full Text | Google Scholar

Elwood, R. W. (1995). The California verbal learning test: psychometric characteristics and clinical application. Neuropsychol. Rev. 5, 173–201. doi:10.1007/BF02214761

PubMed Abstract | CrossRef Full Text | Google Scholar

Engelman, C. D., Koscik, R. L., Jonaitis, E. M., Okonkwo, O. C., Hermann, B. P., La Rue, A., et al. (2013). Interaction between two cholesterol metabolism genes influences memory: findings from the Wisconsin registry for Alzheimer’s prevention. J. Alzheimer’s Dis. 36, 749–757. doi:10.3233/JAD-130482

PubMed Abstract | CrossRef Full Text | Google Scholar

Ester, M., Kriegel, H.-P., Sander, J., and Xu, X. (1996). A density-based algorithm for discovering clusters in large spatial databases with noise. Kdd 96, 226–231.

Google Scholar

Fan, J., Tao, W., Li, X., Li, H., Zhang, J., Wei, D., et al. (2019). The contribution of genetic factors to cognitive impairment and dementia: apolipoprotein e gene, gene interactions, and polygenic risk. Int. J. Mol. Sci. 20, 1177. doi:10.3390/ijms20051177

PubMed Abstract | CrossRef Full Text | Google Scholar

Furney, S., Simmons, A., Breen, G., Pedroso, I., Lunnon, K., Proitsi, P., et al. (2011). Genome-wide association with mri atrophy measures as a quantitative trait locus for alzheimer’s disease. Mol. Psychiatry 16, 1130–1138. doi:10.1038/mp.2010.123

PubMed Abstract | CrossRef Full Text | Google Scholar

Gabryelewicz, T., Religa, D., Styczynska, M., Peplonska, B., Pfeffer, A., Wasiak, B., et al. (2002). Behavioural pathology in alzheimer’s disease with special reference to apolipoprotein e genotype. Dementia Geriatric Cognitive Disord. 14, 208–212. doi:10.1159/000066020

PubMed Abstract | CrossRef Full Text | Google Scholar

Greenacre, M., Groenen, P. J., Hastie, T., d’Enza, A. I., Markos, A., and Tuzhilina, E. (2022). Principal component analysis. Nat. Rev. Methods Prim. 2, 100. doi:10.1038/s43586-022-00184-w

CrossRef Full Text | Google Scholar

Haley, G. E., Berteau-Pavy, F., Park, B., and Raber, J. (2010). Effects of ɛ4 on object recognition in the non-demented elderly. Curr. Aging Sci. 3, (2) 127–137. doi:10.2174/1874609811003020127

PubMed Abstract | CrossRef Full Text | Google Scholar

Hanamsagar, R., and Bilbo, S. D. (2016). Sex differences in neurodevelopmental and neurodegenerative disorders: focus on microglial function and neuroinflammation during development. J. Steroid Biochem. Mol. Biol. 160, 127–133. doi:10.1016/j.jsbmb.2015.09.039

PubMed Abstract | CrossRef Full Text | Google Scholar

Hindley, G., Shadrin, A. A., van der Meer, D., Parker, N., Cheng, W., O’Connell, K. S., et al. (2023). Multivariate genetic analysis of personality and cognitive traits reveals abundant pleiotropy. Nat. Human Behav. 7, 1584–1600. doi:10.1038/s41562-023-01630-9

PubMed Abstract | CrossRef Full Text | Google Scholar

Hirono, N., Hashimoto, M., Yasuda, M., Kazui, H., and Mori, E. (2003). Accelerated memory decline in alzheimer’s disease with apolipoprotein ɛ4 allele. J. Neuropsychiatry Clin. Neurosci. 15, 354–358. doi:10.1176/jnp.15.3.354

PubMed Abstract | CrossRef Full Text | Google Scholar

Holland, D., Desikan, R. S., Dale, A. M., and McEvoy, L. K., and Alzheimer's Disease Neuro, and imaging Initiative (2013). Higher rates of decline for women and apolipoprotein e ɛ4 carriers. Am. J. Neuroradiol. 34, 2287–2293. doi:10.3174/ajnr.A3601

PubMed Abstract | CrossRef Full Text | Google Scholar

Hoogmartens, J., Cacace, R., and Van Broeckhoven, C. (2021). Insight into the genetic etiology of alzheimer’s disease: a comprehensive review of the role of rare variants. Alzheimer’s Dementia Diagnosis, Assess. Dis. Monit. 13, e12155. doi:10.1002/dad2.12155

PubMed Abstract | CrossRef Full Text | Google Scholar

Hsu, Y.-C., Lin, M.-C., Su, M.-H., Cheng, C.-F., Pan, Y.-J., Fan, C. C., et al. (2025). Education as a modifier of genetic influence on cognitive ability in older adults. Behav. Genet. 55, 384–394. doi:10.1007/s10519-025-10229-x

PubMed Abstract | CrossRef Full Text | Google Scholar

Huang, Y., and Mahley, R. W. (2014). Apolipoprotein e: structure and function in lipid metabolism, neurobiology, and alzheimer’s diseases. Neurobiol. Dis. 72, 3–12. doi:10.1016/j.nbd.2014.08.025

PubMed Abstract | CrossRef Full Text | Google Scholar

Huang, Y. A., Zhou, B., Nabet, A. M., Wernig, M., and Südhof, T. C. (2019). Differential signaling mediated by apoe2, apoe3, and apoe4 in human neurons parallels Alzheimer’s disease risk. J. Neurosci. 39, 7408–7427. doi:10.1523/JNEUROSCI.2994-18.2019

PubMed Abstract | CrossRef Full Text | Google Scholar

Hubacek, J. A., Piper, B. J., Pikhart, H., Peasey, A., Kubinova, R., and Bobak, M. (2013). Lack of an association between left-handedness and apoe polymorphism in a large sample of adults: results of the Czech hapiee study. Laterality Asymmetries Body Brain Cognition 18, 513–519. doi:10.1080/1357650X.2012.715164

PubMed Abstract | CrossRef Full Text | Google Scholar

Izaks, G. J., Gansevoort, R. T., van der Knaap, A. M., Navis, G., Dullaart, R. P., and Slaets, J. P. (2011). The association of apoe genotype with cognitive function in persons aged 35 years or older. PLoS One 6, e27415. doi:10.1371/journal.pone.0027415

PubMed Abstract | CrossRef Full Text | Google Scholar

Jack Jr, C. R., Wiste, H. J., Lesnick, T. G., Weigand, S. D., Knopman, D. S., Vemuri, P., et al. (2013). Brain β-amyloid load approaches a Plateau. Neurology 80, 890–896. doi:10.1212/WNL.0b013e3182840bbe

PubMed Abstract | CrossRef Full Text | Google Scholar

Jansen, I. E., Savage, J. E., Watanabe, K., Bryois, J., Williams, D. M., Steinberg, S., et al. (2019). Genome-wide meta-analysis identifies new loci and functional pathways influencing alzheimer’s disease risk. Nat. Genetics 51, 404–413. doi:10.1038/s41588-018-0311-9

PubMed Abstract | CrossRef Full Text | Google Scholar

Jaworowska, A., and Szustrowa, T. (2010). Test matryc ravena w wersji standard forma równoległa. Warszawa: Pracownia Testów Psychologicznych PTP.

Google Scholar

Jochemsen, H. M., Muller, M., van der Graaf, Y., and Geerlings, M. I. (2012). Apoe ɛ4 differentially influences change in memory performance depending on age. The smart-mr study. Neurobiol. Aging 33, 832–e15. doi:10.1016/j.neurobiolaging.2011.07.016

PubMed Abstract | CrossRef Full Text | Google Scholar

Jun, G., Naj, A. C., Beecham, G. W., Wang, L. S., Buros, J., Gallins, P. J., et al. (2011). Meta-analysis confirms cr1, clu, and picalm as alzheimer disease risk loci and reveals interactions with apoe genotypes. Arch. Neurol. 67, (12) 1473–1484. doi:10.1001/archneurol.2010.201

PubMed Abstract | CrossRef Full Text | Google Scholar

Lakens, D. (2022). Sample size justification. Collabra Psychol. 8, 33267. doi:10.1525/collabra.33267

CrossRef Full Text | Google Scholar

Lancaster, C., Tabet, N., and Rusted, J. (2016). The apoe paradox: do attentional control differences in mid-adulthood reflect risk of late-life cognitive decline. Neurobiol. Aging 48, 114–121. doi:10.1016/j.neurobiolaging.2016.08.015

PubMed Abstract | CrossRef Full Text | Google Scholar

Le Carret, N., Lafont, S., Letenneur, L., Dartigues, J.-F., Mayo, W., and Fabrigoule, C. (2003). The effect of education on cognitive performances and its implication for the constitution of the cognitive reserve. Dev. Neuropsychol. 23, 317–337. doi:10.1207/S15326942DN2303_1

PubMed Abstract | CrossRef Full Text | Google Scholar

Li, T., Rezaeipanah, A., and El Din, E. M. T. (2022). An ensemble agglomerative hierarchical clustering algorithm based on clusters clustering technique and the novel similarity measurement. J. King Saud University-Computer Inf. Sci. 34, 3828–3842. doi:10.1016/j.jksuci.2022.04.010

CrossRef Full Text | Google Scholar

Liu, C.-C., Kanekiyo, T., Xu, H., and Bu, G. (2013). Apolipoprotein e and alzheimer disease: risk, mechanisms and therapy. Nat. Rev. Neurol. 9, 106–118. doi:10.1038/nrneurol.2012.263

PubMed Abstract | CrossRef Full Text | Google Scholar

Liu, Y.-B., Wang, X.-J., Tan, L., Tan, C.-C., Xu, W., and Initiative, A. D. N. (2023). Picalm variation moderates the relationships of apoe 4 with alzheimer’s disease cerebrospinal biomarkers and memory function among non-demented population. J. Alzheimer’s Dis. 96, 1651–1661. doi:10.3233/JAD-230516

PubMed Abstract | CrossRef Full Text | Google Scholar

Livingston, G., Huntley, J., Liu, K. Y., Costafreda, S. G., Selbæk, G., Alladi, S., et al. (2024). Dementia prevention, intervention, and care: 2024 report of the lancet standing commission. Lancet 404, 572–628. doi:10.1016/S0140-6736(24)01296-0

PubMed Abstract | CrossRef Full Text | Google Scholar

Łojek, E., Stańczak, J., Delis, D. C., Kramer, J. H., and Ober, B. A. (2010). Podręcznik do kalifornijskiego testu uczenia się językowego CVLT Deana C. Delisa, Joela H. Kramera. Ed. Kaplan I Beth A. Ober Polska Normalizacja Prac. Testów Psychol. Pol. Tow. Psychol.

Google Scholar

Luciano, M., Gow, A. J., Harris, S. E., Hayward, C., Allerhand, M., Starr, J. M., et al. (2009). Cognitive ability at age 11 and 70 years, information processing speed, and apoe variation: the lothian birth cohort 1936 study. Psychol. Aging 24, 129–138. doi:10.1037/a0014780

PubMed Abstract | CrossRef Full Text | Google Scholar

Lundberg, S. M., and Lee, S.-I. (2017). A unified approach to interpreting model predictions. Adv. Neural Information Processing Systems. doi:10.48550/arXiv.1705.07874

CrossRef Full Text | Google Scholar

Mengel-From, J., Christensen, K., McGue, M., and Christiansen, L. (2011). Genetic variations in the clu and picalm genes are associated with cognitive function in the oldest old. Neurobiol. Aging 32, 554–e7. doi:10.1016/j.neurobiolaging.2010.07.016

PubMed Abstract | CrossRef Full Text | Google Scholar

Miozzo, M., Fischer-Baum, S., and Caccappolo-van Vliet, E. (2013). Perseverations in Alzheimer’s disease: memory slips? Cortex 49, 2028–2039. doi:10.1016/j.cortex.2012.10.016

PubMed Abstract | CrossRef Full Text | Google Scholar

Morgen, K., Ramirez, A., Frölich, L., Tost, H., Plichta, M. M., Kölsch, H., et al. (2014). Genetic interaction of picalm and apoe is associated with brain atrophy and cognitive impairment in Alzheimer’s disease. Alzheimer’s Dementia 10, S269–S276. doi:10.1016/j.jalz.2013.11.001

PubMed Abstract | CrossRef Full Text | Google Scholar

Neu, S. C., Pa, J., Kukull, W., Beekly, D., Kuzma, A., Gangadharan, P., et al. (2017). Apolipoprotein e genotype and sex risk factors for alzheimer disease: a meta-analysis. JAMA Neurol. 74, 1178–1189. doi:10.1001/jamaneurol.2017.2188

PubMed Abstract | CrossRef Full Text | Google Scholar

Pakhomov, S. V., Eberly, L. E., and Knopman, D. S. (2018). Recurrent perseverations on semantic verbal fluency tasks as an early marker of cognitive impairment. J. Clin. Experi. Neuropsychol. 40, 832–840. doi:10.1080/13803395.2018.1438372

PubMed Abstract | CrossRef Full Text | Google Scholar

Panico, F., Sagliano, L., Magliacano, A., Santangelo, G., and Trojano, L. (2023). The relationship between cognitive reserve and cognition in healthy adults: a systematic review. Curr. Psychol. 42, 24751–24763. doi:10.1007/s12144-022-03523-y

CrossRef Full Text | Google Scholar

Pettigrew, C., Soldan, A., Li, S., Lu, Y., Wang, M.-C., Selnes, O. A., et al. (2013). Relationship of cognitive reserve and apoe status to the emergence of clinical symptoms in preclinical alzheimer’s disease. Cogn. Neurosci. 4, 136–142. doi:10.1080/17588928.2013.831820

PubMed Abstract | CrossRef Full Text | Google Scholar

Pettigrew, C., Nazarovs, J., Soldan, A., Singh, V., Wang, J., Hohman, T., et al. (2023). Alzheimer’s disease genetic risk and cognitive reserve in relationship to long-term cognitive trajectories among cognitively normal individuals. Alzheimer’s Res. Ther. 15, 66. doi:10.1186/s13195-023-01206-9

PubMed Abstract | CrossRef Full Text | Google Scholar

Pinto, C., and Tandel, K. Y. (2016). Cognitive reserve: concept, determinants, and promotion. J. Geriatric Mental Health 3, 44–51. doi:10.4103/2348-9995.181916

CrossRef Full Text | Google Scholar

Plomin, R., and Deary, I. J. (2015). Genetics and intelligence differences: five special findings. Mol. Psychiatry 20, 98–108. doi:10.1038/mp.2014.105

PubMed Abstract | CrossRef Full Text | Google Scholar

Plomin, R., and Von Stumm, S. (2018). The new genetics of intelligence. Nat. Rev. Genet. 19, 148–159. doi:10.1038/nrg.2017.104

PubMed Abstract | CrossRef Full Text | Google Scholar

Ponomareva, N., Andreeva, T., Protasova, M., Konovalov, R., Krotenkova, M., Malina, D., et al. (2020). Genetic association between alzheimer’s disease risk variant of the picalm gene and eeg functional connectivity in non-demented adults. Front. Neurosci. 14, 324. doi:10.3389/fnins.2020.00324

PubMed Abstract | CrossRef Full Text | Google Scholar

Raulin, A.-C., Doss, S. V., Trottier, Z. A., Ikezu, T. C., Bu, G., and Liu, C.-C. (2022). Apoe in alzheimer’s disease: pathophysiology and therapeutic strategies. Mol. Neurodegeneration 17, 72. doi:10.1186/s13024-022-00574-4

PubMed Abstract | CrossRef Full Text | Google Scholar

Raven, J. (2008). The raven progressive matrices tests: their theoretical basis and measurement model. Uses Abuses Intell. Stud. Advancing Spearman Raven’s Quest Non-arbitrary Metrics, 17–68.

Google Scholar

Raven, J. (2003). Raven progressive matrices. Handbook of nonverbal assessment. Boston, MA: Springer.

Google Scholar

Reinvang, I., Winjevoll, I. L., Rootwelt, H., and Espeseth, T. (2010). Working memory deficits in healthy apoe epsilon 4 carriers. Neuropsychologia 48, 566–573. doi:10.1016/j.neuropsychologia.2009.10.018

PubMed Abstract | CrossRef Full Text | Google Scholar

Religa, D., Styczynska, M., Peplonska, B., Gabryelewicz, T., Pfeffer, A., Chodakowska, M., et al. (2003). Homocysteine, apolipoproteine e and methylenetetrahydrofolate reductase in alzheimer’s disease and mild cognitive impairment. Dementia Geriatric Cognitive Disorders 16, 64–70. doi:10.1159/000070677

PubMed Abstract | CrossRef Full Text | Google Scholar

Ruchinskas, R. (2019). Wechsler adult intelligence scale-digit span performance in subjective cognitive complaints, amnestic mild cognitive impairment, and probable dementia of the Alzheimer type. Clin. Neuropsychologist 33, 1436–1444. doi:10.1080/13854046.2019.1585574

PubMed Abstract | CrossRef Full Text | Google Scholar

Rusted, J., Evans, S., King, S., Dowell, N., Tabet, N., and Tofts, P. (2013). Apoe e4 polymorphism in young adults is associated with improved attention and indexed by distinct neural signatures. Neuroimage 65, 364–373. doi:10.1016/j.neuroimage.2012.10.010

PubMed Abstract | CrossRef Full Text | Google Scholar

Sampedro, F., Vilaplana, E., De Leon, M. J., Alcolea, D., Pegueroles, J., Montal, V., et al. (2015). Apoe-by-sex interactions on brain structure and metabolism in healthy elderly controls. Oncotarget 6, 26663–26674. doi:10.18632/oncotarget.5185

PubMed Abstract | CrossRef Full Text | Google Scholar

Sasirekha, K., and Baby, P. (2013). Agglomerative hierarchical clustering algorithm a review. Int. J. Sci. Res. Publ. 3 (3), 1–3.

Google Scholar

Scheltens, N. M., Galindo-Garre, F., Pijnenburg, Y. A., van der Vlies, A. E., Smits, L. L., Koene, T., et al. (2016). The identification of cognitive subtypes in Alzheimer’s disease dementia using latent class analysis. J. Neurol. Neurosurg. Psychiatry 87, 235–243. doi:10.1136/jnnp-2014-309582

PubMed Abstract | CrossRef Full Text | Google Scholar

Schultz, M. R., Lyons, M. J., Franz, C. E., Grant, M. D., Boake, C., Jacobson, K. C., et al. (2008). Apolipoprotein e genotype and memory in the sixth decade of life. Neurology 70, 1771–1777. doi:10.1212/01.wnl.0000286941.74372.cc

PubMed Abstract | CrossRef Full Text | Google Scholar

Singhal, P., Verma, S. S., and Ritchie, M. D. (2023). Gene interactions in human disease studies–evidence is mounting. Annu. Rev. Biomed. Data Sci. 6, 377–395. doi:10.1146/annurev-biodatasci-102022-120818

PubMed Abstract | CrossRef Full Text | Google Scholar

Styczynska, M., Religa, D., Pfeffer, A., Luczywek, E., Wasiak, B., Styczynski, G., et al. (2003). Simultaneous analysis of five genetic risk factors in polish patients with Alzheimer’s disease. Neurosci. Letters 344, 99–102. doi:10.1016/s0304-3940(03)00438-5

PubMed Abstract | CrossRef Full Text | Google Scholar

Sun, D.-M., Chen, H.-F., Zuo, Q.-L., Su, F., Bai, F., and Liu, C.-F. (2017). Effect of picalm rs3851179 polymorphism on the default mode network function in mild cognitive impairment. Behav. Brain Res. 331, 225–232. doi:10.1016/j.bbr.2017.05.043

PubMed Abstract | CrossRef Full Text | Google Scholar

Sundermann, E. E., Tran, M., Maki, P. M., Bondi, M. W., and Initiative, A. D. N. (2018). Sex differences in the association between apolipoprotein e ɛ4 allele and alzheimer’s disease markers. Alzheimer’s Dementia Diagnosis Assess. Dis. Monit. 10, 438–447. doi:10.1016/j.dadm.2018.06.004

PubMed Abstract | CrossRef Full Text | Google Scholar

Sweet, R. A., Seltman, H., Emanuel, J. E., Lopez, O. L., Becker, J. T., Bis, J. C., et al. (2012). Effect of Alzheimer’s disease risk genes on trajectories of cognitive function in the cardiovascular health study. Am. J. Psychiatry 169, 954–962. doi:10.1176/appi.ajp.2012.11121815

PubMed Abstract | CrossRef Full Text | Google Scholar

Thomas, K. R., Eppig, J., Edmonds, E. C., Jacobs, D. M., Libon, D. J., Au, R., et al. (2018). Word-list intrusion errors predict progression to mild cognitive impairment. Neuropsychology 32, 235–245. doi:10.1037/neu0000413

PubMed Abstract | CrossRef Full Text | Google Scholar

Torres, V. L., Rosselli, M., Loewenstein, D. A., Curiel, R. E., Vélez Uribe, I., Lang, M., et al. (2019). Types of errors on a semantic interference task in mild cognitive impairment and dementia. Neuropsychology 33, 670–684. doi:10.1037/neu0000542

PubMed Abstract | CrossRef Full Text | Google Scholar

Tuminello, E. R., and Han, S. D. (2011). The apolipoprotein e antagonistic pleiotropy hypothesis: review and recommendations. Int. J. Alzheimer’s Disease 2011, 726197. doi:10.4061/2011/726197

PubMed Abstract | CrossRef Full Text | Google Scholar

Vonk, J. M., Arce Rentería, M., Medina, V. M., Pericak-Vance, M. A., Byrd, G. S., Haines, J., et al. (2019). Education moderates the relation between apoe 4 and memory in nondemented Non-hispanic black older adults. J. Alzheimer’s Dis. 72, 495–506. doi:10.3233/JAD-190415

PubMed Abstract | CrossRef Full Text | Google Scholar

Walsemann, K. M., Jackson, H. M., Boardman, J. D., and Herd, P. (2025). Apoe genotype and cognitive decline: educational context as a moderator of genetic risk. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 80, gbaf110. doi:10.1093/geronb/gbaf110

PubMed Abstract | CrossRef Full Text | Google Scholar

Wang, Y., Liu, S., Spiteri, A. G., Huynh, A. L. H., Chu, C., Masters, C. L., et al. (2024). Understanding machine learning applications in dementia research and clinical practice: a review for biomedical scientists and clinicians. Alzheimer’s Res. Therapy 16, 175. doi:10.1186/s13195-024-01540-6

PubMed Abstract | CrossRef Full Text | Google Scholar

Wilczewski, C. M., Obasohan, J., Paschall, J. E., Zhang, S., Singh, S., Maxwell, G. L., et al. (2023). Genotype first: clinical genomics research through a reverse phenotyping approach. Am. J. Hum. Genet. 110, 3–12. doi:10.1016/j.ajhg.2022.12.004

PubMed Abstract | CrossRef Full Text | Google Scholar

Wu, Z., Yang, Y., Song, Z., Ma, M., Feng, M., Liu, Y., et al. (2022). Picalm rs3851179 variants modulate left postcentral cortex thickness, csf amyloid β42, and phosphorylated tau in the elderly. Brain Sci. 12, 1681. doi:10.3390/brainsci12121681

PubMed Abstract | CrossRef Full Text | Google Scholar

Wu, Z., Chen, J., Liu, Y., Yang, Y., Feng, M., Dai, H., et al. (2024). The effects of picalm rs3851179 and age on brain atrophy and cognition along the Alzheimer’s disease continuum. Mol. Neurobiol. 61, 6984–6996. doi:10.1007/s12035-024-03953-8

PubMed Abstract | CrossRef Full Text | Google Scholar

Xu, W., Tan, C.-C., Cao, X.-P., Tan, L., and Initiative, A. D. N. (2020). Association of Alzheimer’s disease risk variants on the picalm gene with picalm expression, core biomarkers, and feature neurodegeneration. Aging (Albany NY) 12, 21202–21219. doi:10.18632/aging.103814

PubMed Abstract | CrossRef Full Text | Google Scholar

Zeng, F.-F., Liu, J., He, H., Gao, X.-P., Liao, M.-Q., Yu, X.-X., et al. (2019). Association of picalm gene polymorphisms with Alzheimer’s disease: evidence from an updated meta-analysis. Curr. Alzheimer Res. 16, 1196–1205. doi:10.2174/1567205016666190805165607

PubMed Abstract | CrossRef Full Text | Google Scholar