Chemical Space Exploration: How Genetic Algorithms Find the Needle in the Haystack

Abstract

We attempt to explain why search algorithms can find molecules with particular properties in an enormous chemical space (ca 10 60 molecules) by considering only a tiny subset (typically 10 3−6 molecules). Using a very simple example, we show that the number of potential paths that the search algorithms can follow to the target is equally vast. Thus, the probability of randomly finding a molecule that is on one of these paths is quite high and from here a search algorithm can follow the path to the target molecule. A path is defined as a series of molecules that have some non-zero quantifiable similarity (score) with the target molecule and that are increasingly similar to the target molecule. The minimum path length from any point in chemical space to the target corresponds is on the order of 100 steps, where a step is the change of and atom-or bond-type. Thus, a perfect search algorithm should be able to locate a particular molecule in chemical space by screening on the order of 100s of molecules, provided the score changes incrementally. We show that the actual number for a genetic search algorithm is between 100 and several millions, and depending on the target property and its dependence on molecular changes, the molecular representation, and the number of solutions to the search problem.

EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation

“…The evolutionary approach on molecular graphs is an efficient method to find needles in a haystack [ 46 ]. Thanks to our sequential and atom centred process, we can also visualise the progression of the exploration, allowing a better interpretation of the results (see Fig.…”

“…One can also note the generally positive effect of secondary actions that worked with half as many steps. Inspired by Henault et al in their exploration of the chemical space with a genetic algorithm, we can look more precisely at our efficiency in the rediscovery tasks [ 46 ]. In Table 4 we note that the secondary actions have a huge impact on the number of calls to the evaluation function in order to find the target.…”

MolFinder: An Evolutionary Algorithm for the Global Optimization of Molecular Properties and the Extensive Exploration of Chemical Space Using SMILES

“…In addition to recent ML-based approaches, various genetic algorithm (GA)-based molecular property optimization algorithms have been developed [25,26,27,28,29,30,31,32,33]. The main advantage of GA-based algorithms is that they do not require a large amount of molecule data relevant to a given optimization task because they search novel molecules in a combinatorial and stochastic way.…”

“…Most existing GA-based molecular optimization algorithms are based on the graph representation of a molecule. In recent studies, they showed competitive, sometimes better, performance compared to ML-based methods in generating novel molecules with desired properties [26,28,29,25]. A GA-based method using the graph representation requires careful design of crossover and/or mutation operations of graphs, which may bias the direction and extension of chemical space search.…”

“…Unlike most existing graph-based GA methods, MolFinder performs a highly efficient global optimization of molecular properties using the SMILES representation. This contradicts the preconception of the field that using the SMILES representation with an evolutionary algorithm is relatively inefficient [26,28,29,31,33]. For the global optimization of molecular properties, MolFinder employs the CSA algorithm, which has been successfully applied to many global optimization problems in various disciplines [35,36,37,38,39].…”