Combining some traditional experimental methods of molecular biology with computational methods of artificial intelligence, a group of researchers from Ruder Boškovic Instititute and Faculty of Natural Sciences and Mathematics from Zagreb, Croatia, demonstrated a novel approach for producing ‘protein fingerprints’ of diverse tissues. This result could lead to the development of new convenient methods in medical diagnostics.
Being directly responsible for a great majority of processes in living cells, proteins are the most important class of biological molecules. They are literally ‘molecular machines’ which facilitate the import of nutrients into the cell and expulsion of waste products from it, production of energy and transportation of material within the cell, as well as cellular respiration and mechanical motion. Due to their immense importance, proteins are among the most vigorously studied topics in biology today.
Over half a million different protein species have been identified in humans, each of them related to particular types of human cells. Different tissues, such as muscles, bones, nerves or skin, are distinguished by the unique ‘protein fingerprint’ – the specific relative abundance of different proteins contained in their cells. Moreover, pathological changes in any type of tissue necessarily have an impact on the tissue’s protein composition, and therefore protein fingerprinting can be used for early diagnostics and identification of various diseases such as tumors or infections.
Unfortunately, producing a good quality protein fingerprint has until now been a complicated, time consuming and expensive enterprise. However, based on their research of tumors conducted on horseradish plant tissue, the Croatian team proposed a novel approach to bypass this obstacle.
Applying computational methods of artificial intelligence, they ‘trained’ a computer to precisely extract the most relevant information on the protein fingerprint from rather ‘fuzzy’ experimental data obtained by 1D protein electrophoresis, a well known, simple, quick and cheap experimental method of molecular biology. Their result hence opens up the possibility for development of a cheap, convenient and reliable method for producing good quality protein fingerprints.
The study 'Enhanced analytical power of SDS-PAGE using machine learning algorithms' will be published in the January issue of “Proteomics” (DOI 10.1002/pmic.200700555)
Cite This Page: