Supporting supervised learning in fungal Biosynthetic Gene Cluster discovery: new benchmark datasets

Abstract

Fungal Biosynthetic Gene Clusters (BGCs) of secondary metabolites are clusters of genes capable of producing natural products, compounds that play an important role in the production of a wide variety of bioactive compounds, including antibiotics and pharmaceuticals. Identifying BGCs can lead to the discovery of novel natural products to benefit human health. Previous work has been focused on developing automatic tools to support BGC discovery in plants, fungi, and bacteria. Datadriven methods, as well as probabilistic and supervised learning methods have been explored in identifying BGCs. Most methods applied to identify fungal BGCs were data-driven and presented limited scope. Supervised learning methods have been shown to perform well at identifying BGCs in bacteria, and could be well suited to perform the same task in fungi. But labeled data instances are needed to perform supervised learning.Openly accessible BGC databases contain only a very small portion of previously curated fungal BGCs. Making new fungal BGC datasets available could motivate the development of supervised learning methods for fungal BGCs and potentially improve prediction performance compared to data-driven methods. In this work we propose new publicly available fungal BGC datasets to support the BGC discovery task using supervised learning. These datasets are prepared to perform binary classification and predict candidate BGC regions in fungal genomes. In addition we analyse the performance of a well supported supervised learning tool developed to predict BGCs.

TOUCAN: a framework for fungal biosynthetic gene cluster discovery

“…DeepBGC achieved a 0.923 area under the curve when predicting BGC positions in a set of 65 experimentally validated BGCs from six bacterial genomes, outperforming previous studies ( 15 ). When handling fungal BGC data, DeepBGC in its original version yielded performance no higher than 0.2 F -m ( 17 ), and when trained on fungal data underperformed previous methods such as fungiSMASH ( 11 ), as we show in the ‘Results’ section. This could indicate that BGC discovery methods developed for bacteria may not be suitable for fungi due to the high diversity of fungal BGCs that are found to vary even among closely related species ( 3 ).…”

“…TOUCAN classification models were developed with comprehensive and exhaustive fungal BGC datasets presented in ( 17 ) that are publicly available to support benchmarking of BGC discovery methods. The six fungal BGC training datasets are composed of different distributions of positive instances obtained from the MIBiG (Minimum Information about a Biosynthetic Gene cluster) ( 2 ) repository and synthetic negative instances generated from OrthoDB ( 18 ) orthologues.…”

“…To build negative instances, the amino acid sequences of OrthoDB fungal orthologous genes were concatenated using a fixed window size of 7000 amino acids, which corresponds to the average amino acid length of all positive instances from the fungal subset in MIBiG. This process generated a pool of training samples of 693 195 synthetic negative clusters [see ( 17 ) for details]. Studying datasets of various distributions could shed light on the impact of class imbalance in fungal BGC discovery, which by nature presents a highly imbalanced scenario where only a small fraction of fungal genomes actually corresponded to BGCs ( 17 ).…”

“…This process generated a pool of training samples of 693 195 synthetic negative clusters [see ( 17 ) for details]. Studying datasets of various distributions could shed light on the impact of class imbalance in fungal BGC discovery, which by nature presents a highly imbalanced scenario where only a small fraction of fungal genomes actually corresponded to BGCs ( 17 ). To account for genomic diversity in fungi, positive instances in the six datasets represent >10 different BGC types and >100 fungal species, while negative instances were generated from a pool of orthologous genes representing ≈300 fungal species.…”

“…In the test phase, we evaluated our classification models with six test datasets, generated similarly to ( 17 ), from a manually curated genome sequence of Aspergillus niger NRRL3, available at https://gb.fungalgenomics.ca/portal . Aspergillus niger is an organism of interest for BGC discovery due to its relevance to industrial processes, and its ubiquitous distribution ( 6 ).…”