PgmNr 193: Integrating molecular and clinical phenotypes towards clinical insight on genotype-phenotype relationships for missense germline PTEN variation.Authors:
S. Thacker 1, 2, 3; T. Mighell 1, 4; I.N. Smith 2; M. Seyfi 2; B.J. O'Roak 5, 9; C. Eng 2, 3, 6, 7, 8, 9
View Session Add to Schedule
1) These authors contributed equally; 2) Genomic Medicine Institute, Lerner Research Institute, Cleveland Clinic, Cleveland, OH 44195, USA; 3) Cleveland Clinic Lerner College of Medicine, Cleveland, OH 44195, USA; 4) Neuroscience Graduate Program, Department of Molecular & Medical Genetics, Oregon Health & Science University, Portland, OR 97239; 5) Department of Molecular & Medical Genetics, Oregon Health & Science University, Portland, OR 97239; 6) Taussig Cancer Institute, Cleveland Clinic, Cleveland, OH 44195, USA; 7) Department of Genetics and Genome Sciences, Case Western Reserve University School of Medicine; Cleveland, OH 44106, USA; 8) Germline High Risk Cancer Focus Group, Comprehensive Cancer Center, Case Western Reserve University School of Medicine; Cleveland, OH 44106, USA; 9) Joint senior authors
PTEN is a nexus between cancer and autism spectrum disorder (ASD), where germline mutations lead to either condition. The single gene-disparate phenotype observation is a clinical challenge that begs resolution. In this study, we sought to integrate molecular phenotype data (i.e. downstream information from all possible variants) from a clinically, rigorously annotated cohort of PTEN Hamartoma Tumor Syndrome (PHTS) individuals (N = 256) in order to differentiate between clinical outcomes (i.e. cancer or ASD). We found that the effect a PTEN missense variant has on the lipid phosphatase activity (i.e. fitness score) of PTEN explains 40% of the variation in disease burden (i.e. Cleveland Clinic Score) and 22% of the variation in head circumference in adult PHTS patients. Furthermore, we found that abundance score (i.e. the steady-state expression of a PTEN mutant) explains 9% of both the variation in disease burden and head circumference in adult PHTS patients. These findings lend critical insight into how variants modulating the lipid phosphatase activity of PTEN explain the penetrance and expressivity of PHTS overgrowth phenotypes. Notably, the above relationships only hold for missense, not nonsense, variation, suggesting that missense variants can participate in different PTEN biology. Despite the clinically informative nature of molecular phenotypes, we found that in isolation, they could not predict clinical outcome; there were no differences in their distributions between PTEN-ASD and PTEN-cancer groups (P>0.05). Subsequently, modeling aggregated molecular phenotype data for pathogenicity and clinical outcomes for missense variants showed high (AUC = 0.92) and moderate accuracy (AUC = 0.77), respectively. The accuracy of these models confirms molecular phenotypes inform clinical outcomes. Finally, we applied unsupervised k-means clustering to identify six distinct groups of PHTS individuals: two ASD clusters, two cancer clusters, and two mixed clusters (between sum-of-squares variance = 50.6%). Together, our data illustrate genotype-specific effects influence clinical outcomes, highlighting the deeply shared biology of ASD and cancer and indicating the existence of divergent points. Obtaining comprehensive molecular phenotype data will inform precision care for PHTS individuals (versus a cohort).