A Typical Kobra Written by Dr Kedar Joshi
A typical and/or ideal Kobra (Kokanastha Brahmin: one from Maharashtra, India) is : White, ruthless, arrogant, egoistic, sharp, intelligent, precise, neat, tidy, stubborn, deep, simple, tolerant (often except with mankind), humorous, ironic, hardworking, determined, passionate, handsome, crazy, An artist, racist, introvert, commanding, optimist, bloody-minded, unromantic, learned, studious, boring, hot & short-tempered, cynic, feminine, practical, sadistic, depressed,
| | Protein Design: Automated protein discovery and synthesisWritten by Paras Chopra
In this paper I describe (theoretically) method(s) of automated protein discovery and synthesis. 1. Protein Folding Problem To solve protein folding problem we can use Artificial Neural Networks. We will train networks with natural proteins whose 3D structure and amino acid sequence is known. After that we will test network with few new artificially designed proteins to check if it works correctly. If it doesn't, we will be changing some of network's parameter such as training iterations, no of hidden layers, etc. And train network again. To check protein's 3D structure, we need to have a model of actual physical world in computer model. 2. Simulation of Physical World This is trickiest part. To simulate physical world at atomic level is very difficult. We need to take into account: covalent bonds, spatial & temporal parameters, weak interactions such as hydrogen bonds, dipole interactions, etc. We also need to simulate chemical reactions. This will probably require huge amounts of computing power. Or perhaps, neural networks can be employed here also as little inaccuracy produced by a neural network can take care of randomness at quantum level. The neural networks will be used to predict/calculate magnitude of effect of various forces on an atom/molecule and also how these behave at a grander inter-molecular level. 3. Designing Proteins To design proteins, we will be using Genetic Algorithm method. The random amino-acid sequences will be evolved & tested by converting these sequences into their respective 3D shape by trained neural network. The best sequences will be retained, while other mutated or crossed-over, etc. The fitness function will work in simulated physical world. If protein produced is successful in carrying out our desired unction, then it is fit else it is not. Actually we will assign a fitness level from 0 to 100. Once final amino acid sequence is determined, it will be sent to Protein Printer.
|