MATHEMATICAL MODELING AND ADVANCED COMPUTING IN DRUG DESIGN AND BIOINFORMATICS
The proposed project is a continuation of the previous research of the Lab of Drug Design and Bioinformatics at the Faculty of Pharmacy of the Medical University of Sofia.
Drug design is a rational approach for discovery of new drug molecules by using in silico methods to analyze the interactions between candidate drugs and human proteins. Drug design is an environmentally friendly and inexpensive approach, as the synthesis of toxic substances and animal experiments are kept to a minimum. In the same time, it is fast and efficient approach because the in silico experiments are much shorter than the in vitro and in vivo experiments.
Bioinformatics explores biology by IT methods like applied mathematics, computer science, statistics, artificial intelligence, machine learning. It is able to model biological systems and functions, analyze and generate models based on big data, predict new outcomes using mathematical models, recognize patterns in experimental data, predict functions of genes and proteins, perform in silico experiments.
The tasks of the Lab defined in the CoE in Informatics and ICT are as follows:
Task 1: IT-based design of novel anti-Alzheimer drugs
We focus on the discovery of novel lead compounds with dual-site binding on acetylcholinesterase (AChE). AChE is one of the main targets in Alzheimer therapy. It controls the cholinergic neurotransmission in the brain and accelerates the assembly of amyloid-beta-peptides into Alzheimer’s fibrils. Recently, in our Lab was performed a virtual screening of ZINC database. The ZINC database contains more than 6 mln known drug and drug-like molecules. The screening discovered 12 novel structures with high affinity to AChE. The structures were synthesized and tested in vitro. Nine of them showed affinities higher than the affinities of the known AChE inhibitors. In the present study, novel hybrids of the newly discovered inhibitors will be designed and screened by molecular docking and molecular dynamics simulations. The best performing molecules will be synthesized and tested in vitro and in vivo.
Task 2: Development of web application for allergenicity prediction
Allergenicity is a subtle, non-linearly coded protein property. We have developed two sequence-based models for allergenicity prediction: AllergenFP and AllerTOP which are freely available at http://www.ddg-pharmfac.net/services. In the proposed task, we will develop a structure-based model for allergenicity prediction. The structures of allergens and non-allergens will be collected from protein families’ databases (Pfam, SCOP) and will be coded by binary descriptor finderprints. A discriminant analysis will be applied to define an allergen fingerprint for each protein family. These fingerprints will be used to predict the allergenicity of novel proteins. The model will be implemented in a web application for allergenicity prediction.
Task 3: Update and upgrade of web applications for immunogenicity prediction
Immunogenicity is the ability of a foreign or self-protein to provoke an immune response in the host. We have developed several web applications for immunogenicity prediction: EpiTOP, EpiDOCK, EpiJen, MHCPred and VaxiJen. As new experimental data for known or new modified or unmodified proteins are continually appearing in the literature, these applications need to be updated and upgraded regularly. In the present study, VaxiJen will be updated and upgraded. VaxiJen is our most widely used application by scientists around the globe. The VaxiJen database will be updated and used to develop new models for immunogenicity prediction based on machine learning methods.