Scaling structural learning with NO-BEARS to infer causal transcriptome networks

TL;DR

This paper introduces NO-BEARS, an efficient and accurate algorithm for inferring gene regulatory networks from transcriptomic data, leveraging GPU acceleration and novel constraints to handle non-linearity and reduce computational costs.

Contribution

The paper presents NO-BEARS, a new algorithm that improves upon NOTEARS by adding a new constraint, a polynomial regression loss, and GPU-based implementation for faster, more accurate network inference.

Findings

Enhanced inference accuracy on synthetic data

Significantly reduced computation time with GPU implementation

Effective handling of non-linear gene expression relationships

Abstract

Constructing gene regulatory networks is a critical step in revealing disease mechanisms from transcriptomic data. In this work, we present NO-BEARS, a novel algorithm for estimating gene regulatory networks. The NO-BEARS algorithm is built on the basis of the NOTEARS algorithm with two improvements. First, we propose a new constraint and its fast approximation to reduce the computational cost of the NO-TEARS algorithm. Next, we introduce a polynomial regression loss to handle non-linearity in gene expressions. Our implementation utilizes modern GPU computation that can decrease the time of hours-long CPU computation to seconds. Using synthetic data, we demonstrate improved performance, both in processing time and accuracy, on inferring gene regulatory networks from gene expression data.