# DyeDactic workflow to predict halochromism of biosynthetic colourants

## TL;DR

This paper introduces a computational workflow to predict how biosynthetic dyes change color with pH, helping design sustainable, eco-friendly textile dyes.

## Contribution

A novel workflow combining machine learning and quantum chemistry to predict halochromism in biosynthetic colorants.

## Key findings

- DyeDactic successfully predicted color changes for emodin, quinalizarin, biliverdin, and orcein.
- The workflow was validated experimentally and applied to understand the color shift of bikaverin during autoclaving.
- The method enables chemoenzymatic modifications for industrial dye applications.

## Abstract

Textile dyeing using microorganisms is a step towards sustainable manufacturing. Computational design offers the prospect of new biosynthetic colourants with better dyeing performance, greater photostability, reduced toxicity, and desired colour. We present a workflow (DyeDactic) to predict halochromism, i.e. colour at different pH values. We filter compound libraries using a graph neural network model to estimate the relevant electronic transition energies of potential colourants. The absorption spectra in the visible region and the colours of the resultant molecules are calculated using time-dependent density functional theory. The populations of protonated and deprotonated species are estimated at different pH values. A weighted sum of their computed absorption spectra gives the predicted colour. The DyeDactic workflow is applied to four natural colourants: emodin, quinalizarin, biliverdin, and orcein, followed by experimental validation. As an illustration we also investigated the molecular mechanism of a red to blue colour change when microbial culture containing polyketide bikaverin is autoclaved. The workflow represents a useful tool to guide chemoenzymatic modifications to achieve industrial applicability.

Biosynthetic colourants are sustainable alternatives to replace artificial dyes, but prediction of their optical properties remains challenging. Here, the authors develop a workflow combining message passing neural networks and excited state calculations to predict colour changes at different pH levels, offering a promising tool for designing eco-friendly dyes.