# Novel TMS-derived metrics enable machine learning classification of major depressive disorder

## TL;DR

This study introduces new TMS-based metrics that, when combined with machine learning, can accurately classify individuals with major depressive disorder.

## Contribution

The paper introduces two novel TMS-derived cortical excitability metrics as potential biomarkers for MDD.

## Key findings

- The novel metrics δ and ϱ significantly improved classification accuracy compared to raw MEPs alone.
- Combining δ, ϱ, and MEPs achieved 83.3% accuracy and 82.3% balanced accuracy in classifying MDD.
- The results suggest δ and ϱ capture neurophysiological changes specific to MDD.

## Abstract

No validated biomarker currently exists for early detection or personalized treatment of major depressive disorder (MDD). Transcranial magnetic stimulation (TMS) is widely used in clinical and research settings and holds promise for biomarker discovery. We assessed two novel TMS-derived cortical excitability metrics, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\delta$$\end{document}δ and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varrho$$\end{document}ϱ, for distinguishing individuals with MDD from healthy controls. Motor-evoked potentials (MEPs) were recorded from the left abductor pollicis brevis during TMS of the right primary motor cortex in twenty-six unmedicated MDD patients and seventeen never-depressed controls. \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\delta$$\end{document}δ and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varrho$$\end{document}ϱ were computed from peak-to-peak MEP amplitudes. A Gradient Boosting classifier predicted diagnostic status using raw MEPs, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\delta$$\end{document}δ and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varrho$$\end{document}ϱ, or their combination. While MEPs alone were non-predictive, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\delta$$\end{document}δ and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varrho$$\end{document}ϱ significantly improved accuracy. Combining MEPs with \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\delta$$\end{document}δ and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varrho$$\end{document}ϱ yielded 83.3% accuracy and 82.3% balanced accuracy. These results suggest \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\delta$$\end{document}δ and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varrho$$\end{document}ϱ effectively capture neurophysiological alterations in MDD and support their potential as candidate biomarkers for MDD.

Diagnosing major depressive disorder (MDD) is challenging because the condition has a wide array of presenting symptoms. Thus, a combination of multiple biomarkers capturing different aspects of the condition is needed. We formulated two novel proxies of cortical excitability using transcranial magnetic stimulation (TMS). Combined with a simple machine learning model, these metrics predicted depression with >82% accuracy. This fills an important gap, since previously proposed biomarkers were not typically based on cortical excitability.