Fmrp targets or not: long, highly brain-expressed genes tend to be implicated in autism and brain disorders

Table 1 A linear model based on transcript expression and length predicts a substantial proportion of Fmrp HITS-CLIP data

Linear model	Fmrp count >1	Fmrp count >16
	n = 7,207 genes	n = 1,228 genes
	r ²	r ²	p
Fmrp count ~ transcript abundance	0.41	0.21	p < 2.2e − 16
Fmrp count ~ Cds length	0.21	0.12	p < 2.2e − 16
Fmrp count ~ transcript length	0.23	0.09	p < 2.2e − 16
Fmrp count ~ abundance + length (either)	0.54	0.44	p < 2.2e − 16
Fmrp count ~ abundance + Cds length + transcript length	0.61	0.44	p < 2.2e − 16

Using either all genes with at least one CLIP read, expression (logCPM) or length (also log2) predicts some of the Fmrp CLIP tag depth (left column). A linear model incorporating all three has an r ² > 0.6. Limiting the analysis only to those genes with high read count (>16), the model still has an r ² > 0.4. All models are highly significant (P < 2.2e − 16).

ISSN: 2040-2392