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

BMC Bioinformatics

Table 1 Comparison of DNAzyme catalytic core sequence descriptors in sequence-only k_obs prediction task

Type	Descriptors	Q²	R²_test	RMSE_test
Bioinformatics-based	2-mers	0.181	0.146	0.942
	3-mers	0.214	0.076	0.980
	4-mers	0.169	0.058	0.985
	one-hot	0.150	−0.138	1.082
Secondary structure-based	encoded dot-bracket	−0.072	0.073	0.969
LLM-based	HyenaDNA + PCA	0.207	0.179	0.923
Property-based	AE embeddings	0.259	0.267	0.871

ISSN: 1471-2105