BACKGROUND: IN ORDER TO DEVELOP A QSAR MODEL, MOLECULAR DESCRIPTORS ARE USED AS INDEPENDENT VARIABLES. NOWADAYS, THOUSANDS OF DESCRIPTORS CAN BE CALCULATED BY MEANS OF DEDICATED SOFTWARE. HOWEVER, WHEN MODELING A PARTICULAR PROPERTY, IT IS REASONABLE TO ASSUME THAT ONLY A SMALL NUMBER OF DESCRIPTORS ARE SUITABLE FOR BUILDING THE MATHEMATICAL MODEL OF INTEREST. AS A CONSEQUENCE, A KEY STEP IS THE SELECTION OF THE OPTIMAL SUBSET OF MOLECULAR DESCRIPTORS FOR THE DEVELOPMENT OF THE MODEL. THIS IS PRECISELY THE AIM OF THE SO-CALLED FEATURE SELECTION METHODS. FEATURE SELECTION METHODS ARE MORE SIGNIFICANT WHEN THE NUMBER OF FEATURES IS ABUNDANT. AMONG THE FEATURE SELECTION METHODS, STOCHASTIC SEARCH ALGORITHMS THAT USE BINARY VERSION OF HEURISTIC SEARCH ALGORITHMS ARE MORE NOTABLE BECAUSE THEY FIND THE NEAR OPTIMUM SOLUTION IN A REASONABLE TIME. SO FAR, MANY ALGORITHMS HAVE BEEN PROPOSED TO OVERCOME FEATURE SELECTION PROBLEM, BUT NONE OF THEM BEHAVE GENERALLY. THEREFORE, PRESENTATION OF NEW FEATURE SELECTION ALGORITHMS IS STILL IMPORTANT. ACCORDINGLY, TWO NEW FEATURE SELECTION ALGORITHMS NAMELY BINARY GRAVITATIONAL SEARCH ALGORITHM (BGSA) AND QUANTUM-BEHAVED PARTICLE SWARM OPTIMIZATION (QPSO) WERE DEVELOPED, CODED AND APPLIED FOR USING IN QSAR STUDIES.METHOD: BGSA IS INTRODUCED BASED ON THE METAPHOR OF GRAVITY AND MOTION LAWS [1]. IN THIS ALGORITHM, THE SEARCHER AGENTS ARE A COLLECTION OF MASSES WHICH CAN DETERMINE THE POSITION AND STATUS OF THE OTHER MASSES VIA GRAVITATIONAL FORCE. THE POSITION OF EACH MASS IS A SOLUTION OF THE PROBLEM. ACCORDING TO THE FITNESS OF EACH AGENT IN BAYESIAN REGULARIZED ARTIFICIAL NEURAL NETWORK (BRANN), A MASS IS ATTRIBUTED. THE MASSES CHANGE THEIR POSITION BASED ON FORCES EXERTED FROM OTHER MASSES. AFTER OPTIMIZING THE ALGORITHM PARAMETERS, SEARCHER AGENTS ARE AGGREGATED IN GLOBAL OPTIMUM AND ALGORITHM IS CONVERGED. THE BGSA WAS APPLIED AS FEATURE SELECTION ALGORITHM FOR INVESTIGATING THE QUANTITATIVE RELATIONSHIP BETWEEN STRUCTURES OF IMIDAZO [4, 5- B] PYRIDINE DERIVATIVES AND THEIR ANTI-CANCER ACTIVITY. QPSO ALGORITHM [2] CONSIDERS THE SEARCH SPACE AS A SYSTEM WITH QUANTUM PARTICLES BY INSPIRATION OF HEISENBERG‟S UNCERTAINTY PRINCIPLE AND THEN SCANS IT. IN FACT, THIS ALGORITHM IS PROBABILITY-BASED VERSION OF PSO ALGORITHM IN WHICH PARTICLES MOVES IN QUANTUM MANNER INSTEAD OF NEWTONIAN MODE. A POTENTIAL WELL ATTRACTS THE PARTICLES RELATIVE TO THEIR FITNESSES. THE FITNESS OF PARTICLES IS DETERMINED BY BRANN. THE PARTICLES TRANSFER TO THEIR NEW POSITIONS AFTER EXCHANGING THEIR INFORMATION WITH EACH OTHER. THIS METHOD WAS SUCCESSFULLY APPLIED FOR ANTI-HIV ACTIVITY MODELING OF FLAVONOID DERIVATIVES.