Quantum annealing versus classical machine learning applied to a simplified computational biology problem

Richard Y Li,Daniel A Lidar,Remo Rohs,Rosa Di Felice

doi:10.1038/s41534-018-0060-8

Richard Y Li, Daniel A Lidar + Show 2 more

Open Access

https://doi.org/10.1038/s41534-018-0060-8

Copy DOI

Abstract

Transcription factors regulate gene expression, but how these proteins recognize and specifically bind to their DNA targets is still debated. Machine learning models are effective means to reveal interaction mechanisms. Here we studied the ability of a quantum machine learning approach to classify and rank binding affinities. Using simplified data sets of a small number of DNA sequences derived from actual binding affinity experiments, we trained a commercially available quantum annealer to classify and rank transcription factor binding. The results were compared to state-of-the-art classical approaches for the same simplified data sets, including simulated annealing, simulated quantum annealing, multiple linear regression, LASSO, and extreme gradient boosting. Despite technological limitations, we find a slight advantage in classification performance and nearly equal ranking performance using the quantum annealer for these fairly small training data sets. Thus, we propose that quantum annealing might be an effective method to implement machine learning for certain computational biology problems.

Highlights

The adiabatic theorem of quantum mechanics, which underlies quantum annealing (QA), implies that a physical system will remain in the ground state if a given perturbation acts slowly enough and if there is a gap between the ground state and the rest of the system’s energy spectrum[25] (Fig. 1a)
QA devices may be of use for studying a simplified biological problem, we report results obtained by solving a learning protocol with six different strategies: (i) an adiabatic quantum machine learning approach formulated in refs. 4,5 (DW), (ii) simulated annealing[48] (SA), (iii) simulated QA50 (SQA), a classical algorithm that can represent the potential of a noiseless thermal quantum annealer, (iv) L2 regularized multiple linear regression (MLR), (v) Lasso[51] and (vi) a scalable machine learning tool known as XGBoost (XGB).[52]
In this work we have explored the possibility of using a machine learning algorithm based on QA to solve a simplified but actual biological problem, the classification and ranking of transcription factor (TF)-DNA binding events

Summary

Introduction

Quantum computation has been the subject of intense scientific scrutiny for its potential to solve certain fundamental problems, such as factoring of integers[1] or simulation of quantum systems,[2,3] more efficiently than classical algorithms, by using unique quantum phenomena including entanglement and tunneling.More recently, there has been much interest in the potential of quantum machine learning to outperform its classical counterparts.[4,5,6,7,8,9,10,11,12,13,14,15,16,17,18] different implementations and models of quantum computing are still in development, promising theoretical and experimental research indicates that quantum annealing (QA),[19] or adiabatic quantum optimization,[20] may be capable of providing advantages in solving classically-hard problems that are of practical interest (for a review see ref. 21). The adiabatic theorem of quantum mechanics, which underlies QA, implies that a physical system will remain in the ground state if a given perturbation acts slowly enough and if there is a gap between the ground state and the rest of the system’s energy spectrum[25] (Fig. 1a). The adiabatic theorem ensures that the ground state of the system at t = tf will give the desired solution to the problem, provided the interpolation is sufficiently slow, i.e., tf is large compared to the timescale set by the inverse of the smallest ground state gap of H(s) and by dH(s)/ds[27] (Fig. 1a)

Objectives

Methods

Results

Conclusion