1. Methods to extract latent information from data.

All available information in experimental data, if present, should be fully embedded in the space determined by both the data themselves and their respective latent prior. For example, in a sequencing experiment, sequenced fragments are the data, and the prior includes the experimental method, the sampled cell type, and the species-specific reference genome. To maximize the recovery of information content from such data, we need methods that minimize information loss during data transformation.

While classical bioinformatic approaches explicitly fits the data into expert-defined models defined by the priors, they segregate information into different domains, which results in information loss. The Zhang laboratory develop novel computational methods to compress the observed data together with its prior, to obtain an approximation of the probability distribution which maximizes the likelihood of such observation. Such tool would enable us to resolve latent biological information from the data with improved signal preservation.

2. Methods to resolve dynamic biological process.

Delineating the history of development and diseases would help us to identify the determining molecular events. However, real-time tracking and locating of molecular events in vivo is always difficult, if not impossible. Most times, what we can obtain are end-point samples instead of time-resolved snapshots of the dynamically transforming process. Furthermore, observation itself does not allow direct manipulation studies to validate findings and distinguish the cause and consequential events.

The Zhang laboratory develop novel experimental and computational methods to track molecular events in time and space, particularly those events previously untraceable by molecular genetics or sequencing methods.

3. Dissect molecular mechanism of epigenome replication.

All cells in an individual multicellular organism share mostly identical genomic sequence. Epigenomic modification enables cell fate determination and subsequent stable inheritance of phenotype within the lineage. Failure in the replication of the parental epigenome to the daughter would result in aging, oncogenesis, and developmental disorders.

The Zhang laboratory track the replication of epigenome between parental and daughter copies to determine its key underlying molecular mechanism. To do this, a combination of experimental and computational methods ranging from perturb-seq, single-molecule imaging, single-molecule sequencing, and deep learning would be used.

2. Provided evidence for how DNA secondary structure affect epigenome replication (BioRxiv, 2024)

3. Developed tools to convert Drosophila neuronal activity into genetic manipulable object (J.Neurogenetics, 2012)