Experimental Confirmation of Multiple Co-Existent DNA Secondary Structures using Low-Yield Bisulfite Sequencing

The prediction of DNA secondary structures from DNA sequences using thermodynamic models is imperfect for many biological sequences, both due to insufficient experimental data for training and to the kinetics of folding that lead to metastable structures. Here, we developed low-yield bisulfite sequencing (LYB-seq) to query the secondary structure states of cytosine (C) nucleotides in thousands of different DNA oligonucleotides on a single-molecule level. We observed that the reaction kinetics between bisulfite and C nucleotides is highly dependent on the secondary structure state of the C nucleotides, with the most accessible C nucleotides (those in small hairpin loops) reacting 70-fold faster than those in stable duplexes. Next, we developed a statistical model to evaluate the likelihood of an NGS read being consistent with a particular proposed secondary structure. By analyzing thousands of NGS reads for each DNA species, we can infer the distribution of secondary structures adopted by each species in solution. We find that 84% of 1,057 human genome subsequences studied here adopt 2 or more stable secondary structures in solution. ### Competing Interest Statement DYZ is a co-founder and significant equity holder of Nuprobe Global, Torus Biosystems, and Pana Bio. JL is a consultant for Torus Biosystems. BY was a full-time employee of Microsoft during this research. MXW is a consultant for Nuprobe USA. JG and AP are full-time employees of Microsoft. There is a patent pending on the low-yield bisulfite sequencing.