genomics intermediate

Mapping variants in thyroid hormone transporter MCT8 to disease severity by genomic, phenotypic, functional, structural and deep learning integration

Stefan Groeneweg, van Geest Ferdy S., Mariano Martin, Mafalda Dias, Jonathan Frazer, Carolina Medina‐Gómez, T. M. Sterenborg Rosalie B., Hao Wang, Anna Dolcetta‐Capuzzo, de Rooij Linda J., Alexander Teumer, Ayhan Abacı, T. Erica L., van den Akker, Gautam P, M. Armour Christine Ambegaonkar, Armour Christine M., Iiuliu Bacos, Priyanka Bakhtiani, Diana Barca, Bauer Andrew J., A. Sjoerd A., M.A. van den Berg, den Berge Amanda van, Enrico Bertini, van der Knoop Marieke M., Nicola Brunetti‐Pierri, Doris Brunner, Marco Cappa, Gerarda Cappuccio, Barbara Castellotti, Claudia Castiglioni, Krishna Chatterjee, Alexander Chesover, Peter Christian, der Spek Jet Coenen-van, M. de Coo Irenaeus F., Regis Coutant, Dana Craiu, Patricia Crock, Goede Christian De, Korcan Demir, Cheyenne Dewey, Alice Dica, Paul Dimitri, G. Dremmen Marjolein H., Rachana Dubey, Anina Enderli, Jan Fairchild, Jonathan Gallichan, Luigi Garibaldi, Belinda George, Gevers Evelien F., Erin Greenup, Annette Hackenberg, Zita Halász, Bianka Heinrich, Hurst Anna C., Tony Huynh, Isaza Amber R., Anna Kłosowska, van der Knoop Marieke M., Daniel Konrad, Koolen David A., Heiko Krude, Abhishek Kulkarni, Alexander Laemmle, LaFranchi Stephen H., Amy Lawson‐Yuen, Jan Lebl, Selmar Leeuwenburgh, M Linder-Lucht, Marti Anna Lopez, Lorea Claudia F., Lourenco Charles M., Lunsing Roelineke J., Greta Lyons, Malikova Jana Krenek, Mancilla Edna E., McCormick Kenneth L., Anne McGowan, Verónica Mericq, Lora Felipe Monti, Carla Moran, Muller Katalin E., Nicol Lindsey E., Isabelle Oliver‐Petit, Laura Paone, Paul Praveen G., Michel Polak, Francesco Porta, Poswar Fabiano O., Christina Reinauer, Klára Roženková, Rowen Seckold, Menevse Tuba Seven, Peter Simm, Anna Simon, Yogen Singh, Marco Spada

Repositori digital de la UPF (Universitat Pompeu Fabra) · Sep 3, 2026 · 1 min read

TL;DR

By integrating genomic, phenotypic, structural, and deep learning data, researchers mapped variants in the MCT8 thyroid hormone transporter to disease severity and developed a classifier that predicts variant pathogenicity (AUC 0.91) and severity (AUC 0.86) for over 8,000 variants.

Problem / question

Predicting the exact phenotypic consequences and severity of genetic variants in rare disorders is challenging, specifically for actionable genes like the X-linked SLC16A2 gene, where loss-of-function variants cause a rare neurodevelopmental and metabolic disorder.

Methods

The authors integrated deep clinical phenotyping, functional assays, structural modeling, and population cohort data from approximately 400,000 individuals. They also developed a deep learning-based pathogenicity-severity classifier to evaluate 8,151 variants of the MCT8 transporter.

Key findings

The study established genotype-phenotype relationships across survival and 24 of 32 disease features for MCT8 deficiency. It identified a mild phenocopy in a population cohort of approximately 400,000 individuals with common MCT8 variants. Structural analysis delineated seven critical functional domains in the transporter. Finally, the deep learning classifier successfully predicted variant pathogenicity with an AUC of 0.91 and disease severity with an AUC of 0.86 across 8,151 variants.

Why it matters

This information-dense mapping strategy provides a generalizable framework for understanding genotype-phenotype relationships, predicting disease severity, and guiding clinical management in other rare genetic disorders.

Limitations

The provided text does not explicitly mention any limitations, caveats, or biases of the study.

Takeaway

Integrating clinical, structural, and population data with deep learning enables highly accurate predictions of disease severity and pathogenicity (AUC 0.91) for thousands of variants in the MCT8 gene.