| Objective: In the recent 20 years,the prevalence of depression was increased rapidly among middle-aged and elderly people.Depression,as a common disorder in the elderly population,is associated with major functional effects,lower quality of life,and a significant risk of suicide.Depression is a complex psychiatric illness and is affected by the interaction among genetic factors,disease factors and environmental factors.Inflammation and oxidative stress play important roles in depression.Many studies have shown that the PRKCA gene is involved in the inflammatory response and oxidative stress process.We speculate that the PRKCA gene may be related to depressive symptoms,but no studies have explored the association between the PRKCA gene and depressive symptoms.Therefore,this study investigated the effects of PRKCA gene single nucleotide polymorphism(SNP)and PRKCA weighted gene score reflecting the overall level of genes on depressive symptoms in the elderly and investigated the multiplicative interaction between PRKCA weighted gene score and environmental factors such as diabetes,fruit intake and gastrointestinal disease on depressive symptoms.In addition,the support vector machine and BP neural network model were constructed and verified.Finally,sensitivity analysis was used to verify the robustness of the research results.Methods: In this study,the elderly aged ≥ 60 in a community in Qingdao were selected by cluster sampling method.A field investigation was carried out by questionnaires,patient health questionnaire-9(PHQ-9),blood samples,etc.A total of 170 Han-unrelated individuals with PHQ-9 scores ≥ 5 were selected as the depressed group.A total of 340Han-unrelated individuals with PHQ-9 scores < 5 from the same community were selected as the non-depressed group by 1:2 matching according to gender and age.According to information such as Hap Map or the database of the Thousand People Genome Project,rs4790904、rs1005651、rs11079657、rs2227857、rs2228945、rs2286674、rs228883、rs61687889、rs7342847、rs7342969、rs9909004 of PRKCA gene were selected as target SNPs.The Chi-square test and Fisher’s exact probability method were used to compare the frequency distribution of genotype and allele of target SNPs loci in two groups.To investigate the relationship between PRKCA gene SNPs and depressive symptoms,the Odd Ratio(OR)and 95% confidence intervals(CI)of additive,codominant,dominant,recessive and over-dominant genetic models of 11 SNPs were calculated by 1:2 paired logistic regression analysis.To check for the presence of linkage disequilibrium and construct haplotype domains,we adopted linkage disequilibrium and haplotype analysis.Calculated PRKCA weighted gene score and explored its correlation with depressive symptoms.The effect of multiplicative interaction on depressive symptoms was investigated by the product term of PRKCA weighted gene score and environmental factors.The support vector machine and neural network model were constructed to predict the model,and the neural network model was used to rank the importance of influencing factors of depressive symptoms in the elderly.Finally,the sensitivity analysis was carried out with the E value.Results: A total of 510 subjects were included after 1:2 matching according to gender and age,including 88 males and 252 females in the non-depressed group and 44 males and 126 females in the depressed group.The age of the non-depressed group was 68.56±6.58 years old and the age of the depressed group was 68.59±6.86 years old.1.Elementary case analysis of the research objects: The results indicated that there were differences between the non-depressed group and the depressed group in diabetes(P=0.001),fruit intake(P=0.030)and gastrointestinal diseases(P=0.007).2.Hardy-Weinberg equilibrium tests: The genotype of 11 SNP loci in both groups all followed Hardy-Weinberg equilibrium(P>0.05),and this indicated that the sample has good group representation.3.Genotypic and allelic comparisons of PRKCA genes: The results showed that the genotype and allele distribution of eleven SNPs of the PRKCA gene had no difference in the two groups(P>0.05).4.The relationship between 11 SNPs loci of PRKCA gene and depressive symptoms:After adjusting for marital status,educational level,smoking status,alcohol consumption,hypertension,diabetes,cardiovascular disease,vegetable intake,fruit intake,and gastrointestinal disease,none of the 11 SNPs loci was statistically associated with depressive symptoms under five genetic models: additive,codominant,dominant,recessive,and over-dominant.5.Haplotype analysis and linkage disequilibrium: The results showed that rs1005651 and rs228883,rs7342969 and rs7342847 were in strong linkage disequilibrium(D’>0.75).A total of two haplotype domains were constructed,and Block 1 consisted of rs1005651 and rs228883 loci.The total frequency of AC and CT haplotypes was ≥ 5%.The frequencies of AC and CT haplotypes were 0.740 and 0.248 in the non-depressed group and 0.726 and0.250 in the depressed group,respectively.Block 2 consisted of rs7342969 and rs7342847 loci.The total frequency of AT,AC and GC haplotypes was ≥ 5%.The frequencies of AT,AC and GC haplotypes were 0.610,0.280,0.107 in the non-depressed group and 0.629,0.285,0.086 in the depressed group,respectively.There was no statistical difference in the frequency of haplotype distribution between the two groups(P>0.05),and the results remained stable after 1000 replacement tests.6.Correlation between PRKCA gene score and depressive symptoms: After adjustment for all covariates,the results of multivariate 1:2 paired logistic regression showed that there was a significant association between PRKCA weighted gene score and depressive symptoms(OR=0.647,95%CI: 0.422-0.990,P=0.045).7.Multiplicative interaction analysis of PRKCA weighted gene score and environmental factors: The multiplicative interaction between PRKCA weighted gene score and diabetes mellitus,fruit intake,and gastrointestinal disease was not statistically significant in both univariate and multivariate models.8.Support vector machine: Participants were randomly divided into a training set(353participants)and a validation set(151 participants)in a 7:3 ratio.The area under the ROC curve of the model was 0.69,indicating good prediction accuracy.9.Neural network model: The area under the ROC curve of the model was 0.718,indicating that the model had good accuracy in predicting the occurrence of depressive symptoms.After standardizing the importance of the input layer factors in the model,the top three were PRKCA gene score,fruit intake and diabetes.10.Sensitivity analysis E value: The E value is used to assess the potential impact of unmeasured confounding factors on the causal conclusions in observational studies.The larger the E value,the stronger the evidence supporting the existence of a causal relationship.In this study,the E value was 1.79 after adjusting all covariables.Conclusion:(1)In this study,no statistically significant association was found between five genetic models of rs4790904、rs1005651、rs11079657、rs2227857、rs2228945、rs2286674、rs228883、rs61687889、rs7342847、rs7342969、rs9909004 and depressive symptoms in the elderly.(2)PRKCA weighted gene score was associated with depressive symptoms in the elderly,and the results were robust after E-value analysis.(3)BP neural network model showed that PRKCA weighted gene score,fruit intake and diabetes were the top three factors affecting depressive symptoms compared with other factors in this study. |
PDF Full Text Request