Diagnosing migraine from genome-wide genotype data: a machine learning analysis - Report - MDSpire

Diagnosing migraine from genome-wide genotype data: a machine learning analysis

  • By

  • Antonios Danelakis

  • Tjaša Kumelj

  • Bendik S Winsvold

  • Marte Helene Bjørk

  • Parashkev Nachev

  • Manjit Matharu

  • Dominic Giles

  • Erling Tronvik

  • Helge Langseth

  • Anker Stubberud

  • The International Headache Genetics Consortium

  • Verneri Anttila

  • Ville Artto

  • Andrea C Belin

  • Anna Bjornsdottir

  • Gyda Bjornsdottir

  • Dorret I Boomsma

  • Sigrid Børte

  • Mona A Chalmer

  • Daniel I Chasman

  • Bru Cormand

  • Ester Cuenca-Leon

  • George Davey-Smith

  • Irene de Boer

  • Martin Dichgans

  • Tonu Esko

  • Tobias Freilinger

  • Padhraig Gormley

  • Lyn R Griffiths

  • Eija Hämäläinen

  • Thomas F Hansen

  • Aster V E Harder

  • Heidi Hautakangas

  • Marjo Hiekkala

  • Maria G Hrafnsdottir

  • M Arfan Ikram

  • Marjo-Riitta Järvelin

  • Risto Kajanne

  • Mikko Kallela

  • Jaakko Kaprio

  • Mari Kaunisto

  • Lisette J A Kogelman

  • Espen S Kristoffersen

  • Christian Kubisch

  • Mitja Kurki

  • Tobias Kurth

  • Lenore Launer

  • Terho Lehtimäki

  • Davor Lessel

  • Lannie Ligthart

  • Sigurdur H Magnusson

  • Rainer Malik

  • Bertram Müller-Myhsok

  • Carrie Northover

  • Dale R Nyholt

  • Jes Olesen

  • Aarno Palotie

  • Priit Palta

  • Linda M Pedersen

  • Nancy Pedersen

  • Matti Pirinen

  • Danielle Posthuma

  • Patricia Pozo-Rosich

  • Alice Pressman

  • Olli Raitakari

  • Caroline Ran

  • Gudrun R Sigurdardottir

  • Hreinn Stefansson

  • Kari Stefansson

  • Olafur A Sveinsson

  • Gisela M Terwindt

  • Thorgeir E Thorgeirsson

  • Arn M J M van den Maagdenberg

  • Cornelia van Duijn

  • Maija Wessman

  • Bendik S Winsvold

  • John-Anker Zwart

  • May 6, 2025

  • 0 min

Share

Machine Learning Enhances Migraine Identification Using Genome-Wide Genotyping

Overview

This study applied machine learning models to genome-wide genotyping data from over 43,000 individuals to classify migraine cases versus controls. Machine learning approaches outperformed traditional polygenic risk scoring, achieving higher accuracy and revealing novel genetic pathways related to migraine.

Background

Migraine is a prevalent neurological disorder with a complex genetic basis, estimated to have around 50% heritability. Genome-wide association studies have identified numerous risk loci, but these explain only a fraction of the heritability, leaving a gap known as missing heritability. Traditional polygenic risk scores assume additive genetic effects and do not capture gene-gene interactions, which may contribute to this missing heritability. Machine learning models can potentially capture both additive and interactive genetic effects to improve migraine risk prediction.

Data Highlights

Dataset Size (Variants)Best ModelHold-out Test AUCPRS AUCStatistical Significance vs PRS
108Light Gradient Boosting Machine0.630.52-0.59P < 0.001 to 0.02
7,771Light Gradient Boosting Machine0.630.52-0.59P < 0.001 to 0.02
7,840Light Gradient Boosting Machine0.630.52-0.59P < 0.001 to 0.02
140,467Multinomial Naïve Bayes0.620.52-0.59P < 0.001 to 0.02

Key Findings

  • Machine learning models achieved an area under the curve (AUC) of approximately 0.62-0.63 in classifying migraine cases versus controls, outperforming polygenic risk scoring (AUC 0.52-0.59).
  • The light gradient boosting machine performed best on smaller variant datasets, while multinomial naïve Bayes was optimal for the largest dataset.
  • Machine learning identified many known migraine-associated genes and pathways from GWAS, validating its biological relevance.
  • Novel pathways related to signal transduction, neurological function, botulinum toxin mechanisms, and calcitonin gene-related peptide receptor were uniquely detected by machine learning.
  • Results suggest migraine genetics involve non-additive and interactive effects, which are better captured by complex machine learning models than by additive polygenic risk scores.

Clinical Implications

Incorporating machine learning approaches into genetic risk assessment for migraine may improve diagnostic accuracy beyond traditional polygenic risk scores. Understanding interactive genetic effects could guide the development of targeted therapies, especially those involving pathways like calcitonin gene-related peptide signaling. Future models with larger datasets may further enhance precision medicine strategies for migraine management.

Conclusion

Machine learning models leveraging genome-wide genotyping data provide superior classification of migraine compared to additive polygenic risk scores, uncovering complex genetic interactions and novel biological pathways. These findings advance understanding of migraine genetics and hold promise for improved predictive and therapeutic approaches.

References

  1. Original Article 2024 -- Identifying Migraine Through Genome-Wide Genotyping Data: An Analysis Using Machine Learning Techniques

Original Source(s)

Diagnosing migraine from genome-wide genotype data: a machine learning analysis

  • by Antonios Danelakis, Tjaša Kumelj, Bendik S Winsvold, Marte Helene Bjørk, Parashkev Nachev, Manjit Matharu, Dominic Giles, Erling Tronvik, Helge Langseth, Anker Stubberud, The International Headache Genetics Consortium, Verneri Anttila, Ville Artto, Andrea C Belin, Anna Bjornsdottir, Gyda Bjornsdottir, Dorret I Boomsma, Sigrid Børte, Mona A Chalmer, Daniel I Chasman, Bru Cormand, Ester Cuenca-Leon, George Davey-Smith, Irene de Boer, Martin Dichgans, Tonu Esko, Tobias Freilinger, Padhraig Gormley, Lyn R Griffiths, Eija Hämäläinen, Thomas F Hansen, Aster V E Harder, Heidi Hautakangas, Marjo Hiekkala, Maria G Hrafnsdottir, M Arfan Ikram, Marjo-Riitta Järvelin, Risto Kajanne, Mikko Kallela, Jaakko Kaprio, Mari Kaunisto, Lisette J A Kogelman, Espen S Kristoffersen, Christian Kubisch, Mitja Kurki, Tobias Kurth, Lenore Launer, Terho Lehtimäki, Davor Lessel, Lannie Ligthart, Sigurdur H Magnusson, Rainer Malik, Bertram Müller-Myhsok, Carrie Northover, Dale R Nyholt, Jes Olesen, Aarno Palotie, Priit Palta, Linda M Pedersen, Nancy Pedersen, Matti Pirinen, Danielle Posthuma, Patricia Pozo-Rosich, Alice Pressman, Olli Raitakari, Caroline Ran, Gudrun R Sigurdardottir, Hreinn Stefansson, Kari Stefansson, Olafur A Sveinsson, Gisela M Terwindt, Thorgeir E Thorgeirsson, Arn M J M van den Maagdenberg, Cornelia van Duijn, Maija Wessman, Bendik S Winsvold, John-Anker Zwart

  • May 6, 2025

Related Content

Who Decides When a Doctor Steps Back?

Cognitive decline, credentialing, and why “I’ll know” often isn’t enough.

Migraine, comorbidity, and risks of severe maternal and neonatal morbidity or mortality: a population-based cohort study

IHS, ACP Differ on Migraine Tx

Experts highlight differences between IHS and ACP recommendations as newer therapies reshape acute and preventive migraine care