Functional genomic hypothesis generation and experimentation by a robot scientist

Abstract

The question of whether it is possible to automate the scientific process is of both great theoretical interest and increasing practical importance because, in many scientific areas, data are being generated much faster than they can be effectively analysed. We describe a physically implemented robotic system that applies techniques from artificial intelligence to carry out cycles of scientific experimentation. The system automatically originates hypotheses to explain observations, devises experiments to test these hypotheses, physically runs the experiments using a laboratory robot, interprets the results to falsify hypotheses inconsistent with the data, and then repeats the cycle. Here we apply the system to the determination of gene function using deletion mutants of yeast (Saccharomyces cerevisiae) and auxotrophic growth experiments. We built and tested a detailed logical model (involving genes, proteins and metabolites) of the aromatic amino acid synthesis pathway. In biological experiments that automatically reconstruct parts of this model, we show that an intelligent experiment selection strategy is competitive with human performance and significantly outperforms, with a cost decrease of 3-fold and 100-fold (respectively), both cheapest and random-experiment selection.

Representation of research hypotheses

“…There is still strong belief that only human intelligence is capable of production of research hypotheses. The Robot Scientist project has proved that a machine can not only automatically generate scientifically valuable hypotheses, but also test them and make conclusions about their validity [14]. …”

“…gene G has function A, drug D can cure disease E, etc. The discovery of such scientific knowledge is based on abductive and deductive inferences, and modern technology is now able to automate the inference of possible new facts and their experimental confirmation [14]. The techniques for inductive inference are also in place, e.g.…”

“…Adam abductively hypothesizes new facts about yeast functional biology by inferring what is missing from a model. In our original work on robot scientists, we used a purely logical approach to hypothesis formation based on applying abductive logic programming to a logical model of a yeast metabolism subset [14]. Unfortunately, this general method is too inefficient to deal with large bioinformatic models.…”

“…Instances of the hypotheses defined in LABORS, and instances of the recorded research units, are stored in a publicly available database [11,13]. All the hypotheses discussed below have been automatically generated by the Robot Scientist “Adam” (A Discovery Machine) [14]. An explicit semantically defined representation of hypotheses enables improvements in the representation of investigations designed to test those hypotheses, and in the consistency and validity checking of the conclusions about those hypotheses within the investigations.…”

“…Our experiences in formally recording research hypotheses, and analysing automatically generated hypotheses are summarised as “lessons learned”. The examples in this paper are from the investigations run by Adam, the investigation into re-discovery of gene functions in aromatic amino acids pathway, and the investigation into novel biological knowledge, which are reported in [14] and [11]; all the information about the investigations, including procedures and data, is available at the Robot Scientist project web site [13]. …”