# A Singapore-centric Fungal Dataset of 518 Cultivated Strains with Visual Phenotypes and Taxonomic Identity

## TL;DR

This paper introduces a dataset of 518 fungal strains from Singapore, capturing their visual traits and taxonomy to support future research on fungal diversity and bioactive compounds.

## Contribution

The paper presents a novel cultivation workflow and dataset for tropical fungi, enabling visual phenotype to taxonomy mapping using computer vision.

## Key findings

- A dataset of 518 diverse fungal strains from Singapore was created, including both terrestrial and marine sources.
- The dataset includes high-resolution imaging and taxonomic data, supporting visual phenotype analysis.
- The workflow enables systematic strain preparation and genomic analysis for underexplored tropical fungi.

## Abstract

The fungal kingdom represents a greatly untapped resource to produce a wide range of bioactive secondary metabolites, including antibiotics, anticancer agents, industrially significant dyes and enzymes. To-date, it is estimated only less than 5% of all fungi have been characterised, a deficit that is especially pronounced in tropical regions like Singapore, where fungal diversity remains underexplored compared to northern hemisphere counterparts. This underlines the urgency and importance of our research which motivated the creation of our curated dataset, aiming to address this gap and contribute to understanding the broader ecosystem. We developed a generalisable cultivation workflow that enables systematic strain preparation, supports high-resolution imaging, and yields sufficient fungal biomass amenable for genomic analyses. This resulted in a diverse collection of 518 phylogenetically and ecologically varied fungal strains from both terrestrial and marine environments in biodiverse Singapore. The curated dataset from this project captures both taxonomic identity and colony-level morphological traits serving as a foundation for visual phenotype to taxonomy mapping through the integration of computer vision.