An Efficient Targeted Drug Design Method Using Framework of Multiscale Encoder-Decoder.

Area of Science:

• Computational biology
• Drug discovery
• Bioinformatics

Background:

• Targeted drug design requires effective integration of complex gene and protein information.
• Existing methods may not fully leverage multilevel features from genetic and amino acid sequences.

Purpose of the Study:

• To develop a novel targeted drug design framework integrating gene and protein features.
• To improve the validity, novelty, and binding affinity of designed drugs.

Main Methods:

• A multiscale encoder-decoder framework for targeted drug design.
• Multilevel Gene Feature Extraction (MGFE) for gene expression data (base and codon features).
• Multi-Embedding Protein Feature Extraction (MPFE) for amino acid sequences, considering structural characteristics.
• Gated Recurrent Unit (GRU) based drug decoder for feature integration and drug generation.

Main Results:

• The proposed method effectively fuses gene and protein information for drug design.
• MGFE simulates gene transcription and translation for feature extraction.
• MPFE extracts protein features without location encoding, leveraging amino acid properties.
• Experimental results demonstrate superior performance over existing methods in validity, novelty, and binding affinity.

Conclusions:

• The developed framework offers a promising approach for targeted drug discovery.
• The integration of MGFE and MPFE enhances the extraction of relevant biological features.
• The GRU-based decoder successfully generates novel targeted drugs with improved characteristics.