Design And Development Of Foundational Library(MORT) For CADD

With the development of computer-aided drug design(CADD) and computational biology, a variety of programs or software packages have been developed, and AMBER is one of the most famous programs. AMBER is composed of several different parts with corresponding functions, for example, XLEAP can be used to manipulate the structure of a molecule and generate the topology and parameter files for molecular simulations. XLEAP is written in C, but it tried to use the object-oriented programming(OOP) paradigm that is not natively supported by C. Therefore, the code of XLEAP is awkward and extremely hard to be extended. It is expected that a program package with similar functions to XLEAP can be developed based on new programming diagram and used as the foundation for the development of new software packages and tools employed in CADD and computational biology. Therefore, in this thesis, we designed and developed a new foundational library, MORT(Molecular Objects and Relevant Templates), which can be used in CADD and computational biology.The basic structure of MORT is described in the first part of this thesis, including the data structure and algorithms. The functions that can be used to handle molecular structure and calculate energy were developed by applying the data structure and algorithms. The functions in MORT can be roughly divided into two categories. The first category includes the functions to delete, modify and add the objects, such as atoms, residues and so on. The other category contains the functions to manipulate the properties of objects, for example, to calculate the energy and to generate the topology files. MORT is written in C++, and employs the relational model instead of the hierarchical model to store data, which makes MORT easier to be understood and extended.Then, based on MORT, an algorithm called Fixbond was developed to predict the bond types of a given molecule. Bond types are related to the force field parameters for molecular simulations, and therefore, it is important to set bond types correctly. There are three main rules in this algorithm for the automatic bond type perception: 1. Hard rules; 2. Length rules; 3. Conjugation rules. The combination of the above three rules makes our algorithm more accurate than the other programs. It performs well in handling structures with false coordinates, but there still remain some challenges, especially for molecules with misconnected and false connected bonds. So we plan to improve our algorithm in two directions in the near future: 1. We plan to design and employ better rules to detect the bonds that should be connected but are ignored by our algorithm and to reduce the redundant bonds that are over-connected; 2. Define the entropy of bond by using more subtle schemes. To achieve this, a training process based on a lot of data is necessary and some parameters and formula should be changed accordingly.