Identification of driver genes for severe respiratory response to COVID-19 via a Quantum Support Vector Machine

A Quantum Support Vector Machine (qSVM) is a quantum adaptation of the classical SVM that can be used for classification designed to be run with a quantum annealer (QA) [1]. In this work we implement a qSVM algorithm on the genomic information samples of ~100 young patients with confirmed COVID19, which are classified into three groups: 1) Healthy control, 2) Oxygen: those that need oxygen, and 3) ARDS: Acute Respiratory Distress Syndrome. The computational task is to classify the data as Healthy control vs ARDS patients or Oxygen vs ARDS. Since the QA samples from the quantum distribution, it retains both the lowest energy solution and some of the next lowest-energy solutions. Because of the suboptimal solutions, we expect qSVM to perform worse on the training data than cSVM. However, suboptimal solutions can generate different decision boundaries. As such, a suitable combination of the suboptimal solutions in qSVM might outperform cSVM on the test data. We demonstrate an improved classification performance for qSVM over cSVM on some test data sets, where the optimal solution of qSVM contains both the lowest energy and some of the excited states solutions. We also use the most informative features of different classifiers as input for structural causal modeling to identify potential driver genes for a severe respiratory response. This identified ADAM9 as the key driver gene, which was experimentally validated [2].

 

[1] Willsch, D., et al. (2020). Comput. Phys. Commun. 248, 107006.

[2] Identification of driver genes for critical forms of COVID-19 in a deeply phenotyped young patient cohort. R. Carapito et al., Science Translational Medicine, https://www.science.org/doi/10.1126/scitranslmed.abj7521 (published Oct. 26, 2021).