Identifying functional SNVs that map to non-coding regions of the genome and alter RNA Structure

Genome-wide a ssociation s tudies ( GWAS) o ften id entify d iseaseassociated m uta-tions i n i ntergenic a nd non-coding r egions of t he g enome. Given the high percentage of the genome that is transcribed, we postulate that for some observed associations the disease phenotype is caused by a structural rearrangement i n a r egulatory r egion of t he R NA t ranscript. To identify s uch mutations w e ha ve pe rformed a genome wide a nalysis o f a ll known di seaseassociated Single Nucleotide Variants (SNVs) from the Human Gene Mutation Database ( HGMD) t hat m ap t o t he untranslated r egions ( UTRs) o f a g ene. Rather t han us ing m inimum free e nergy a pproaches ( e.g. mFold), w e us e a partition f unction c alculation th at ta kes in to c onsideration th e ensemble o f possible R NA c onformations f or a g iven s equence. F or s ix disease-states (Hyperferritinaemia Cataract Syndrome, β-Thalassemia, C artilage-Hair Hypoplasia, Retinoblastoma, Chronic Obstructive Pulmonary Disease (COPD), and Hypertension) we identified multiple SNVs in UTRs that alter the mRNA structural en semble o f t he as sociated g enes. U sing a B oltzmann s ampling procedure f or s ub-optimal R NA s tructures, w e ar e ab le t o ch aracterize and visualize t he n ature o f t he co nformational ch anges i nduced b y t he diseaseassociated mutations in th e s tructural en semble. W e o bserve e xperimentally using chemical probing that SNV induced conformational changes analogous to those i n bacterial r egulatory R iboswitches w hen s pecific l igands bi nd occur readily in the human genome. We propose that the UTR and SNV combinations we id entify constitute a “ RiboSNitch,” th at is a r egulatory R NA in which a specific SNV has a s tructural consequence that results in a d isease phenotype. We find that there are specific structural features, in particular whether the SNV increases or de -creases t he R iboSNitch en semble en tropy t hat ar e g ood predictors of function.