The Genetic Origin of Uneven Cognitive Profiles in Heritable Neurodevelopmental Conditions and Individual Differences: Computational Investigations.

While the heterogeneity and co-occurrence of heritable neurodevelopmental conditions such as autism, attention deficit hyperactivity disorder (ADHD), and dyslexia remain issues of debate, these conditions are nevertheless all characterised by uneven cognitive profiles exhibiting strengths and weaknesses. There have been advances in understanding neural markers and genetic predictors of these conditions, but little insight into how DNA variation can influence functional brain development in such a way as to produce uneven cognitive profiles as developmental outcomes. Uneven cognitive profiles (e.g., across verbal and non-verbal intelligence) also characterise individual differences, and similarly, their genetic basis is little understood. Two main sources of uneven profiles appear possible: that there are regional genetic effects on brain development that act on mechanisms which play an influential role in the development of a particular cognitive or socioemotional ability (domain-specificity); or that genetic effects on brain development have a more widespread influence on neurocomputational properties, but the development of particular abilities is differentially sensitive to variation in those properties (domain-relevance). In this article, we present computational simulations that combine genetic algorithms and artificial neural networks to explore the second of these possibilities, domain relevance. Selection is used to alter the population frequency of alleles that influence the neurocomputational properties of a common substrate, under a polygenic model. Different regions of the substrate become specialised for developing different functions, modelled by five tasks. Across 20 generations, we assess how selection for a given task, which serves to tune substrate-wide neurocomputational properties in favour of this task, serves to alter the development of the other four tasks, which must employ the same range of neurocomputational properties. We demonstrate that such selection can enhance or impair acquisition in non-selected domains, depending on the computational demands of each task domain. We also show that behavioural deficits associate with an increase in the heritability of individual differences. We discuss the results in the context of contemporary theories of the influence of genetic variation on functional and structural brain development, and assess the merits of the domain-specific and domain-relevant accounts of uneven cognitive profiles in neurodevelopmental conditions and individual differences. SUMMARY: Heritable neurodevelopmental conditions such as dyslexia, attention deficit hyperactivity disorder, autism, developmental language disorder, and developmental coordination disorder are characterised by uneven cognitive profiles. However, little is known about how genes produce such uneven cognitive profiles. It is a puzzle because genetic effects on brain development are typically more widespread than areas showing functional specialisation in adults. The work presents computational modelling to demonstrate how the relationship between cognitive domains and processing properties of a substrate constrains behavioural development. A common substrate for different domains (such as association cortex in a cortical lobe), where domains are specialised to different regions with shared properties, may exhibit uneven cognitive profiles if the processing properties of the substrate are better tuned to supporting some domains than others (so-called domain-relevance). Simulations then test the hypothesis that domain relevance is a plausible mechanism for explaining how common genetic variation might contribute to specific and uneven cognitive profiles seen in neurodevelopmental conditions. The computational simulations therefore give insight into how conditions such as dyslexia may emerge, why they would be heritable, and why the relationship between genotype and phenotype in common neurodevelopmental conditions is likely to be highly polygenic.