Gilmore, J. H. | bioRxiv Science

bioRxiv2025

Impacts of perinatal factors on white matter outcome at 8 to 10 years by diffusion tensor imaging

BackgroundWhile perinatal factors are known to influence brain development, their long-term impact on white matter microstructure remains incompletely understood. Previous studies using tract-based spatial statistics (TBSS) have shown limited associations between neonatal measures and later white matter development. MethodsWe investigated associations between perinatal factors (birth weight [BW], gestational age [GA], and head circumference [HC]) and white matter microstructure in 117 children aged 8-10 years from the UNC Early Brain Development Study cohort. Diffusion tensor imaging (DTI) data were analyzed using a fiber tract-based framework examining 54 major white matter tracts. Statistical analysis was performed using a functional analysis of fiber tract profiles. ResultsGA and BW showed widespread significant associations with white matter microstructure (38 and 36 out of 54 tracts, respectively), while HC showed limited associations (3 out of 54 tracts). Post-hoc univariate analysis revealed stronger associations with axial diffusivity (AD) compared to radial diffusivity (RD) or fractional anisotropy (FA). AD associations with BW, GA, and HC were found in 30, 31, and 8 tracts, respectively. ConclusionsUsing a fiber tract-based analysis approach, we demonstrated that GA and BW are strongly predictive of white matter organization at school age, while HC showed limited predictive power. The predominant associations with AD suggest these perinatal factors primarily influence axonal organization rather than myelination. These findings enhance our understanding of how early life factors impact long-term brain development.

neuroscience↗

bioRxiv2022

TwinEQTL: Ultra Fast and Powerful Association Analysis for eQTL and GWAS in Twin Studies

We develop a computationally efficient alternative, TwinEQTL, to a linear mixed-effects model (LMM) for twin genome-wide association study (GWAS) data. Instead of analyzing all twin samples together with LMM, TwinEQTL first splits twin samples into two independent groups on which multiple linear regression analysis can be validly performed separately, followed by an appropriate meta-analysis-like approach to combine the two non-independent test results. Through mathematical derivations, we prove the validity of TwinEQTL algorithm and show that the correlation between two dependent test statistics at each single-nucleotide polymorphism (SNP) are independent of its minor allele frequency (MAF). Thus the correlation is constant across all SNPs. Through simulations, we show empirically that TwinEQTL has well controlled type I error with negligible power loss compared to the gold-standard linear mixed effects models. To accommodate eQTL analysis with twin subjects, we further implement TwinEQTL into a R package with much improved computational efficiency. Our approaches provide a significant leap in terms of computing speed for GWAS and eQTL analysis with twin samples.

genetics↗