A2Sign: Agnostic Algorithms for Signatures-a universal method for identifying molecular signatures from...

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Molecular signatures are crucial for cell-type proportion inference from bulk transcriptomics.
• Current methods require prior biological knowledge, limiting their use with less-studied tissues.
• A data-driven approach is needed to discover cell types and generate signatures from complex biological samples.

Purpose of the Study:

• To introduce A2Sign (Agnostic Algorithms for Signatures), a novel computational framework for identifying cell-type-specific molecular signatures.
• To enable cell-type deconvolution from bulk transcriptome data in arbitrary tissues without prior biological knowledge.
• To account for inter-individual variability and reduce collinearity in molecular signatures.

Main Methods:

• Utilized non-negative tensor factorization (NTF) as the core strategy within the A2Sign framework.
• Developed a global approach applicable to diverse tissues and transcriptomic data types (microarray, RNA-seq).
• Implemented the analysis using annotated Python notebooks and the NMTF package.

Main Results:

• Successfully identified cell-type-specific molecular signatures using a data-driven, agnostic approach.
• Demonstrated the ability to reduce collinearities and incorporate inter-individual variability.
• Generated two novel molecular signatures for deconvoluting up to 16 immune cell types.

Conclusions:

• A2Sign provides a powerful, flexible framework for molecular signature discovery and cell-type deconvolution.
• The method overcomes limitations of knowledge-based approaches, expanding transcriptomic analysis to new biological contexts.
• The developed signatures facilitate accurate immune cell profiling from bulk tissue data.