SequenceCraft: machine learning-based resource for exploratory analysis of RNA-cleaving deoxyribozymes.

Area of Science:

• Biochemistry
• Computational Biology
• Molecular Biology

Background:

• DNAzymes are catalytic DNA molecules with potential in gene therapy and diagnostics.
• Existing DNAzyme discovery methods are inefficient and costly, with low success rates.
• There is a need for in silico approaches to discover novel RNA-cleaving DNAzyme cores with improved efficacy.

Purpose of the Study:

• To develop an open-source, machine learning-based platform for DNAzyme exploratory analysis.
• To enable quantitative estimation of observed rate constants (kobs) for DNAzyme activity.
• To facilitate statistical and clustering analysis of DNAzyme data.

Main Methods:

• Developed SequenceCraft, a machine learning platform for DNAzyme analysis.
• Created a curated database of over 350 RNA-cleaving catalytic cores.
• Implemented property-based sequence representations for conventional and modified nucleotides.
• Utilized an optimized kobs prediction algorithm achieving Q2 > 0.9.

Main Results:

• SequenceCraft enables quantitative kobs estimation and data analysis for DNAzymes.
• The platform supports analysis of both conventional and chemically modified nucleotides.
• The kobs prediction algorithm demonstrates high accuracy on existing experimental data.

Conclusions:

• SequenceCraft significantly advances DNAzyme research by providing an efficient in silico discovery tool.
• The platform offers an alternative to traditional experimental methods, accelerating DNAzyme development.
• This work facilitates the discovery of novel RNA-cleaving DNAzymes with superior catalytic activities.