Although currently focusing on the following directions, but we are always open to new ideas!

De Novo Modification Detection  
We use a graph convolutional network model to predict nanopore sequencing signals from nucleotide chemical structures. Such a framework predicts novel modifications by generalizing chemical groups among known nucleotides, e.g. recapitulating 5-methylcytosine by combining cytosine pyrimidine ring and thymine 5-methyl group.

Genome/Transcriptome-Wide Modification Landscape Determination  
We perform genome-wide native DNA and transcriptome-wide native RNA nanopore sequencing to reveal modification landscapes related to various biological questions, e.g. determine genome-wide 5mC landscape alterations during lung development and reveal aberrant 6mA sites in type-2 diabetic transcriptome.

In Silico Perturbation  
We use a variational autoencoder model to predict gene expression alterations under chemical and genetic perturbations. We further use the in silico perturbation analysis to guide the design of stem cell directed differentiation stategies.