Structure-enhanced graph meta learning for few-shot gene regulatory network inference.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene regulatory network (GRN) inference is crucial for understanding biological regulation.
• Existing deep learning methods often demand substantial labeled data, limiting their applicability.
• Data scarcity poses a significant challenge in biological network inference.

Purpose of the Study:

• To develop a novel model for efficient GRN inference with limited data.
• To address the limitations of current deep learning approaches in data-scarce scenarios.
• To present Meta-TGLink, a structure-enhanced graph meta-learning model for few-shot GRN inference.

Main Methods:

• Formulated GRN inference as a link prediction task.
• Employed a structure-enhanced graph meta-learning framework (Meta-TGLink).
• Combined graph neural networks with Transformer architectures to integrate relational and positional information.

Main Results:

• Meta-TGLink demonstrated superior performance in few-shot GRN inference.
• The model effectively captures transferable regulatory patterns.
• Achieved improved predictive accuracy under data-scarce conditions, outperforming state-of-the-art baselines.

Conclusions:

• Meta-TGLink significantly reduces the dependence on extensive labeled datasets for GRN inference.
• The model shows particular strength in cross-domain few-shot learning scenarios.
• This approach advances the field of computational biology by enabling robust GRN inference with limited data.