ARTIFICIAL INTELLIGENCE AS ROOTS OF BIOINFORMATICS

Student Contributor: Shagufta Quazi, B.Sc. Biotechnology, 3^rd year, Adamas University

In the 1960s, where the health informatics originate, computer algorithms become very useful in processing and handling a large quantity of data. With the availability of numerous types of Artificial Intelligence algorithms, it has become popular among researchers as they can easily explore more information understandably, annotated with description, precision assessments, and explanations. The National Institute of Health (NIH) claims that precision medicine is the best approach for preventing and curing diseases but with the aid of powerful supercomputers and innovative algorithms, it can be more comprehensible.

Bioinformatics is an interdisciplinary field comprising mainly genetics, molecular biology, computer science, mathematics, and statistics. The most basic issues involve modeling biological mechanisms at the molecular level and interpreting results from gathered information. Typically, a bioinformatics approach includes the following steps:

• Gathering information from biological evidence.
• Creating a database of computations.
• Solving a problem regarding computational modeling.
• Testing and assessing computer algorithms.

Artificial intelligence (A.I.) can be referred to as a computer imitating human thinking and behavior so that they can perform some tasks ordered by humans. Researchers have been using AI technologies tremendously, for classification of biological sequences, identification of biological entities, determining characteristics and this includes gathering, organizing, and evaluating large quantities of data, which is beyond human beings.

AI can be a big help to bioinformatics, in some fields like:

• Generative Modelling for Protein Structures: Since it is impossible to compare and learn about true data distribution, the generative models tends to learn the true interpretation of the distribution of data, so that the distributed data that has been generated seems similar to the original one. In this case, GANs (Generative Adversarial Networks) are used, which involves the pairing of a generator (learns to give output) with a discriminator (learns to compare true data from the output). So, these two are neural networks in which the generator attempts to produce the actual image and the discriminator attempts to decide if the image generated is false or true. For eg. GANs are used to produce protein structure and the incomplete pieces of compromised protein structure are predicted. The data includes the use of 3D protein structure which is interpreted as the 2D pairwise alpha-carbon size. 

• DNA Sequencing: DNA sequencing involves determining the order of nucleotide sequences (A, T, G, and C) in a DNA strand, but due to the involvement of massive data crunching, complete DNA sequencing was not feasible, as every organism has its unique nucleotide sequences. Companies such as Deep Genomics uses artificial intelligence to help researchers to understand genetic variation. Particularly, algorithms are built based on patterns found in broad genetic data sources which are then converted into computer models to help people know how key cellular processes influence genetic variation. Next-Generation Sequencing has encompassing new DNA sequencing methods, allowing researchers to sequence an entire human genome in a single day relative to the traditional Sanger sequencing process, which took completion over a decade when the human genome was first sequenced.

• Protein Classification: Proteins include polypeptides and polypeptides include simple amino acid chains. Such chains form a functional protein by folding into the final three-dimensional structure. Proteins are classified into various groups according to their biological function. Although most proteins have a similar primary structure and common evolutionary origin, hence it’s hard to classify proteins. One approach is to create a computer program that compares the unspecified amino acid sequence with the recognized protein sequences, and returns the desired protein classification. Appropriate protein analysis and recognition are of crucial significance since they are essential for much of an organism’s main functions.

• Knowledge Discovery in Biological Databases: Knowledge Discovery from Databases (KDD) is a developing field incorporating database processing methods, artificial intelligence, and statistics. It has been recognized that computational methods are required to combine and analyze heterogeneous knowledge to select genetic variations and properties of functional interactions that have a beneficial influence on the biological outcome of the whole organism. Artificial Intelligence can help bring additional insight into an ever-increasing and voluminous biological information. There is already a large body of biological data available, and their successful use involves the retrieval of valuable data.

• Computer-Aided Drug Design (CADD): Computer-aided drug design uses statistical methods for the identification, creation, and study of drugs and related biologically active molecules. CADD is highly based on IT, databases, and computing tools.  AI can successfully handle those activities. The CADD may be used by simulated screening or the stage of lead optimization at different stages of drug development such as target recognition.

With numerous types of AI algorithms available, it has become popular for researchers to use programs that can identify and exploit their datasets. Scientists need strategies that coherently access the data, annotated with meaning, precise estimations, and descriptions. In bioinformatics, artificial intelligence (AI) can be used to model biological data as well as to make new findings.