Back

Whole Genome Sequencing Illuminates the Developmental Signatures of Human Language Ability

Koomar, T.; Casten, L.; Thomas, T. R.; Koh, J.-Y.; Hofamann, D.; Thenuwara, S.; Momany, A.; O'Brien, M.; Murray, J. C.; Tomblin, J. B.; Michaelson, J. J.

2021-11-24 psychiatry and clinical psychology
10.1101/2021.11.22.21266703 medRxiv
Show abstract

1Language is the foundation of human social interaction, education, commerce, and mental health. The heritability underlying language is well-established, but our understanding of its genetic basis -- and how it compares to that of more general cognitive functioning-- remains unclear. To illuminate the language-specific contributions of rare and common variation, we performed whole genome sequencing in N=350 individuals who were characterized with seven latent language phenotypes. We conducted region, gene, and gene set-based analyses to identify patterns of genetic burden that disproportionately explained these language factors compared to nonverbal IQ. These analyses identified language-specific associations with NDST4 and GRIN2A, with common variant replication of NDST4 in an independent sample. Rare variant burden analyses revealed three distinct functional profiles of genes that make contributions to language: a prenatally-expressed profile with enrichment for chromatin modifiers and broad neuropsychiatric risk, a postnatal cortex-expressed profile with enrichment for ion channels and cognitive/neuropsychiatric associations, and a postnatal, subcortically-expressed profile with enrichment of cilium-related proteins. Compared to a profile strongly associated with nonverbal IQ, these language-related profiles showed less intolerance to damaging variation, suggesting that the selection patterns acting on language differ from patterns linked to intellectual disability. Furthermore, we found evidence that rare potential reversions to an ancestral state are associated with poorer overall specific language ability. The breadth of these variant, gene, and profile associations suggest that while human-specific selection patterns do contribute to language, these are distributed broadly across numerous key mechanisms and developmental periods, and not in one or a few "language genes".

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.