# Supervised and unsupervised deep learning-based approaches for studying DNA replication spatiotemporal dynamics

## TL;DR

This paper introduces deep learning methods to study DNA replication patterns in cells, enabling large-scale analysis of replication dynamics and potential applications in disease research.

## Contribution

The novel contribution is the development of supervised and unsupervised deep learning approaches to detect and classify DNA replication dynamics in mouse embryonic stem cells.

## Key findings

- Supervised machine learning successfully classified S-phase patterns in wild-type mouse embryonic stem cells.
- An unsupervised method detected altered replication dynamics in Rif1-deficient cells and cyclin E overexpression models.
- The methods showed potential applicability to patient samples for studying pathogenic processes.

## Abstract

In eukaryotic cells, DNA replication is organised both spatially and temporally, as evidenced by the stage-specific spatial distribution of replication foci in the nucleus. Despite the genetic association of aberrant DNA replication with numerous human diseases, the labour-intensive methods employed to study DNA replication have hindered large-scale analyses of its roles in pathological processes. In this study, we employ two distinct methodologies. We first apply supervised machine learning, successfully classifying S-phase patterns in wild-type mouse embryonic stem cells (mESCs), while additionally identifying altered replication dynamics in Rif1-deficient mESCs. Given the constraints imposed by a classification-based approach, we then develop an unsupervised method for large-scale detection of aberrant S-phase cells. Such a method, which does not aim to classify patterns based on pre-defined categories but rather detects differences autonomously, closely recapitulates expected differences across genotypes. We therefore extend our approach to a well-characterised cellular model of inducible deregulated origin firing, involving cyclin E overexpression. Through parallel EdU- and PCNA-based analyses, we demonstrate the potential applicability of our method to patient samples, offering a means to identify the contribution of deregulated DNA replication to a plethora of pathogenic processes.

When tested on well-characterised cellular models, supervised and unsupervised deep learning closely recapitulate expected differences in DNA replication dynamics across genotypes, holding promise for identifying aberrant DNA replication at scale.