Prediction of new candidate proteins and analysis of sub-modules and protein hubs associated with seed development in rice (Oryza sativa) using an ensemble network-based systems biology approach

Abstract

Background: Rice is a critical global food source, but it faces challenges due to nutritional deficiencies and the pressures of a growing population. Understanding the molecular mechanisms and protein functions in rice seed development is essential to improve yield and grain quality. However, there is still a significant knowledge gap regarding the key proteins and their interactions that govern rice seed development. Protein-protein interaction (PPI) analysis is a powerful tool for studying developmental processes like seed development, though its potential in rice research is yet to be fully realized. With the aim of unraveling the protein interaction landscape associated with rice seed development, this systems biology study conducted a PPI network-based analysis. Using a list of known seed development proteins from the Gene Ontology (GO) knowledgebase and literature, novel candidate proteins for seed development were predicted using an ensemble of network-based algorithms, including Majority Voting, Hishigaki Algorithm, Functional Flow, and Random Walk with Restart, which were selected based on their popularity and usability. The predictions were validated using enrichment analysis and cross-checked with independent transcriptomic analysis results. The rice seed development sub-network was further analyzed for community and hub detection.

Results: The study predicted 196 new proteins linked to rice seed development and identified 14 sub-modules within the network, each representing different developmental pathways, such as endosperm development and seed growth regulation. Of these, 17 proteins were identified as intra-modular hubs and 6 as inter-modular hubs. Notably, the protein SDH1 emerged as a dual hub, acting as both an intra-modular and inter-modular hub, highlighting its importance in seed development PPI network stability.

Conclusions: These findings, including the identified hub proteins and sub-modules, provide a better understanding of the PPI interaction landscape governing seed development in rice. This information is useful for achieving a systems biology understanding of seed development. This study implements an ensemble of algorithms for the analysis and showcases how systems biology techniques can be applied in developmental biology.

Keywords: Agronomics; Hub proteins; Network-based algorithms; PPI network; Rice; Sub-modules.

Fig. 1

The bioinformatics pipeline used for prediction and validation of protein candidates for seed development in rice. The hub protein and sub-pathway analysis procedures are also represented

Fig. 2

Precision-Recall Curves (A) and Receiver Operator Characteristic Curves (B) for the five algorithms, including Majority Voting (MV), Hishigaki Algorithm (HA), Random Walk with Restart (RWR), and Functional flow (FF), and the ensemble model. The ensemble model outperforms all the other algorithms on both metrics based on its higher area under the curve values

Fig. 3

Modified precision top-N curve for protein predictions related to rice seed development, obtained from the rice protein–protein interaction (PPI) network. The x-axis represents the number of top-ranked predicted proteins (N), while the y-axis shows the modified precision, calculated as the percentage of known differentially expressed proteins (DEPs) among the top N predictions. The curve peaks at N = 454, where a maximum precision of 43.17% was achieved

Fig. 4

A visual representation of the seed development protein–protein interaction (PPI) sub-network of rice, with different colors assigned to the detected sub-modules, based on the Louvain community detection algorithm. The sub-modules are numbered and the most relevant Gene Ontology-Biological Process (GO-BP) term related to seed development according to enrichment analysis results are listed in the legend

Fig. 5

Network visualization of hub proteins (highly connected proteins within the network) in the seed development protein–protein interaction (PPI) sub-network of rice. Node size and color indicate the type of hub classification: larger green nodes, such as Ehd2, represent intra-modular hubs (proteins highly connected within a single module), identified using the Z-score method; larger blue nodes, such as SODCP, represent inter-modular hubs (proteins with high connectivity between modules), identified using the partition coefficient (PC). The SDH1 protein, which functions as both an intra- and inter-modular hub, is shown as a larger orange node. Smaller white nodes represent non-hub proteins (proteins with low connectivity). Sub-modules, identified through the Louvain community detection algorithm, are numbered, and the most significantly enriched Gene Ontology-Biological Process (GO-BP) terms derived from functional enrichment analysis are listed in the legend