Development and validation of a machine learning algorithm prediction for dense granule proteins in Apicomplexa.

Area of Science:

• Parasitology
• Genomics
• Bioinformatics

Background:

• Apicomplexa are pathogenic protists that invade host cells, residing within parasitophorous vacuoles (PVs).
• Dense granule proteins (GRAs) secreted by these parasites are crucial for modifying the PV, but many remain unidentified.
• Understanding GRAs is vital for controlling apicomplexan infections.

Purpose of the Study:

• To develop and apply a novel computational model for identifying potential dense granule proteins (GRAs) in Apicomplexa.
• To experimentally validate predicted GRAs in Neospora caninum.

Main Methods:

• A multi-view attention graph convolutional network (MVA-GCN) model was developed using machine learning and genomic data.
• The MVA-GCN model predicted candidate GRAs in Neospora caninum.
• CRISPR/Cas9 gene editing was used to verify predicted GRAs and create knockout strains.

Main Results:

• The MVA-GCN model successfully screened Neospora caninum, identifying two novel GRAs: NcGRA64a and NcGRA64b.
• Gene tagging confirmed the identification of NcGRA64a and NcGRA64b.
• Knocking out the NcGRA64(a,b) genes did not affect Neospora caninum's in vitro growth, replication, or in vivo virulence.

Conclusions:

• The MVA-GCN deep learning model is effective for discovering dense granule proteins (GRAs) within Apicomplexa genomic datasets.
• This prediction model shows promise for identifying other functional proteins in apicomplexan parasites.