Proteins are large biomolecules that consist of one or multiple chains of amino acid residues. They play vital roles in biological function in the biosystem, including transcription, DNA replication, transporting, etc. It is believed that the structure determines the function. Here we explores the mechanism of functional binding and disease-related self-assembly of biomolecules using energy landscape theory from three aspects: Tau aggregation, PU.1 transcription process, and Heme in protein.

First, we used energy landscape theory to investigate the mechanisms of tau aggregation. Tau isoforms, involved in Alzheimer's and Pick's disease, often aggregate to amorphous phase separation and firbilization. The free energy distributions of these Tau isoforms showed two channels in the aggregation: one leading to more ordered amyloid fibrils and the other non-fibrillar channels leading to an amorphous phase. The different structural properties of the species in the two channels suggest that the interconversion between the two channels will act as a backtrack, suggesting that the two channels are kinetically independent.

Then, we focus on the indirect readout role in PU.1 DNA recognition. The accurate recognition of the binding site is essential for transcription. Many proteins recognize the binding region through "indirect readout in the transcription factor family." The indirect readout is subtle and acts as a long-range function. Here, we investigate the origin of the binding specificity of the PU.1. A nonspecific electrostatically driven DNA mechanism mainly achieves the binding specificity of PU.1. The electrostatic interactions between PU.1 and DNA cause both changes in elastic properties of the DNA and complex DNA conformational/dynamics. PU.1 affects the elastic properties of DNA through second-order mechanical effects. When PU.1 binds to the non-binding region of the DNA, the DNA becomes stiffer. Then, PU.1 then slides along the DNA, which becomes softer as the protein recognizes its specific binding site. This recognition process is driven by configurational entropy effects, suggesting a general mechanism for indirect readout.

At last, we focus on heme proteins, where heme act as an active center of proteins. Here, we explore the role of the heme in protein folding and protein structure computationally. First, we model heme proteins using a hybrid model, which combines the AWSEM Hamiltonian for the protein, and coarse-grained forcefield, with AMBER Hamiltonian for the heme, an all-atom forcefield. Here, we show that both heme b and heme c improve protein structure predictions' accuracy. In the folding process, coordinated covalent bonds for both heme b and heme c drive the heme toward the native pocket. Thioester covalent bonds also drive heme c to the binding pocket. In addition, electrostatics also helps to search binding sites.

We explore the mechanism of functional binding and disease-related self-assembly of biomolecules using energy landscape theory from these three aspects.