This symposium describes aspects of a systems biology approach to intelligence differences and age-related cognitive change being conducted at the University of Edinburgh. The participant substrates for the work are resources such as the Lothian Birth Cohorts (follow-up studies of the Scottish Mental Surveys of 1931 and 1947) and locally-administered studies such as Generation Scotland. The methods used include psychometrics, epidemiology, and various ‘omics approaches. Ian Deary outlines the levels of the systems biology approach to intelligence. He describes the uses of and some results from the newer ‘omics measures of the Lothian Birth Cohorts of 1921 (LBC1921; N = 550) and 1936 (LBC1936; N = 1091), selecting from whole genome sequencing, longitudinal DNA methylation, longitudinal gene expression, stem cells, and post-mortem brain tissue. The next three presentations address, respectively, structural brain imaging, metabolomic, and genetic approaches to differences in intelligence and cognitive change.
Simon cox, using the LBC1936, presents data on the brain structural hallmarks of: i) individual differences in the level of intelligence in older age (cross-sectional data at mean age 73 years); ii) successful cognitive ageing from age 11 to 73 years; and iii) successful ageing within older age (longitudinal data at ages 73 N = 728, 76 N = 488, and 79 years N = 388). When correcting significant associations between intelligence and cortical metrics at age 73 for intelligence at age 11 (as an index of lifetime cognitive ageing), no associations between intelligence and cortical thickness were significant (r < |0.087|). Conversely, more successful cognitive ageing was associated with selectively larger frontal, temporal, and parietal volumes and surface areas (r < |0.180|, FDR q < 0.05). We also report new latent growth curve analyses testing correlated changes in intelligence and cortical metrics across the 8th decade of life.
Stuart Ritchie describes how metabolomics can investigate the chemical ‘fingerprints’ left behind by metabolic processes. We used the BIOCRATES array to measure the levels of around 200 blood metabolites from 593 LBC1936 participants at mean age 73 years. We derived a measure of general intelligence from thirteen cognitive tests, and predicted its variation using three variants of penalised regression (LASSO, Ridge, and Elastic Net) on the metabolomic predictors. Depending on the regression technique, the predictors could explain between 4 and 6 percent of the variance in general intelligence in a hold-out (testing) sample. Functional analysis is ongoing to understand the variables uncovered in feature-selection analyses, and discover how they may mechanistically be associated with brain and cognitive differences.
David Hill uses data from LBC1936 and Generation Scotland to address two big questions regarding the genetics of intelligence, viz. ‘why are family-based estimates of heritability greater than single nucleotide polymorphisms (SNP) based estimates?’ and ‘what genes are associated with intelligence, and what do they do?” First, using two methods to examine the contribution made to intelligence by genetic variants that are poorly captured by genotyped SNPs, and applying them to the 20,000 participants of Generation Scotland, the heritability estimates based on SNP data match those derived using a family-based method. Second, by pooling genome-wide association study (GWAS) data on ~240,000 participants from UK Biobank and international consortia, 187 regions of the genome were found to be associated with intelligence. The resulting analysis was also able to indicate that genes involved with neurogenesis and myelination were associated with intelligence.
Biological & Psychopharmacology , Genetics , Neuroimaging
