Beyond DNA: ML-Empowered Nanopore Base-Calling of 12-Letter Genetic Alphabets

The standard 4-letter genetic code (A, T, G, C) is the blueprint of life on earth. However, beyond this foundational framework lies a realm of artificial genetics, which has now expanded the genetic code up to 12-letter (A, T, G, C, B, S, P, Z, X, K, J, V). Strikingly, at a time when detection methods for genomics and transcriptomics have progressed to their “fourth generation” with successful commercialization, the field of artificial genetics is still in its nascent stage, the “zeroth generation”. Herein, in the framework of DFT and machine learning (ML), we report a next-generation solid-state nanopore sequencing to both assess and decode the DNA code with expanded alphabets. For assessing, we leverage the ML regression tools, which predict the transmission signatures of each natural and xenonucleobase with low mean squared error as validated through DFT. Parameterizing SMILES (simplified molecular input line entry system) strings of expanded alphabets, including isomers, allows the structural, molecular, and bonding configuration of nucleobases to be meticulously incorporated during predictions. Further, custom ML classification tools are developed, and each standard, isoG/isoC, hachimoji, and supernumerary code is decoded with SHAP (Shapley Additive exPlanations) explainability. By introducing ML accelerated nanopore sequencing of supernumerary DNA, we pave the way for rapid analysis of expanded alphabets, offering insights into life’s possibilities across the cosmos.