A plethora of computational approaches can be used to overcome the limitations of experimental techniques. Computational tools have become critical for the integration, representation and visualization of heterogeneous genomics, proteomics and biomedical data 
. Experimental techniques, like yeast-two-hybrid, have enabled to pair-wisely screen protein-protein interactions 
. Nevertheless, the study of protein complex data involving more than two partners is relatively restricted due to the limitations of the currently available high-throughput techniques. Computational approaches complement experimental methods for the detection of protein complexes using protein interaction data. The study of protein interaction networks is important not only from a theoretical point of view, but also has strong practical applications towards the development of new drugs, which could specifically interrupt or modulate protein interactions 
Yeast has several features making it an ideal model to study, not only human disorders, but also the effect of nutraceuticals in the prevention or progress of a disease 
. Oxidative damage has long been considered as a primary threat for neurons, both in neurodegenerative disorders and aging. Free radicals, which among others are produced during normal metabolism, can trigger a series of events that disturb important aspects of the normal cellular function, including enzymatic activity, protein folding, transcription, ion channel activity, transporter function and other processes. Such damage may contribute to a broad range of diseases in the nervous system.
We presented a hypothesis-driven approach to elucidate the role of a small model antioxidant molecule and understand how the yeast cell tunes the flux of intermediates through metabolic routes and restructures the cellular transcriptome and proteome in the presence of such a compound. The phenome and metabolome data obtained from our well-controlled yeast cultivations clearly reflected the presence of an antioxidant compound −demonstrating once again that the systematic use of this simple eukaryotic organism can uncover important features of nutraceutical compounds.
By using this external stimulus and network biology tools, we identified a small, tightly connected sub-network reflecting the biological signature of the yeast cell during stress, and we identified the FMP43 protein −which has not been previously functionally characterized− as an important player in the network architecture. In silico
analysis of FMP43
transcriptional regulation and prediction of the post-translational modifications of the corresponding protein revealed a putative new cell cycle regulatory gene. This hypothesis was verified by the significant improvement of the specific growth rate of the yeast cell after deletion of FMP43
and complements the recent finding about cell cycle delay phenotypes observed by over-expression of FMP31 
, a similar protein with as well unknown function.
However, the linkage between antioxidant compounds and a growth-controlling gene (FMP43) needs further investigation. The ProtFun 2.2 server predicts FMP43 as a protein involved in oxidative energy metabolism, possibly due to its role in the metabolism of reactive oxygen species, which correlates well with the high metabolic necessity when time between cell division cycles is longer. This scenario could also explain the decrease in FMP43 expression levels upon addition of the antioxidant molecule and the dual role of this protein in yeast.
Proteins can bind with many types of molecules using a wide variety of binding sites. For example, binding sites used by natural ligands or substrates, allosteric regulatory sites used by products or reversible/irreversible inhibitors, and ‘special’ binding sites at which an array of compounds −such as drugs and antioxidants− can bind 
. Changes in the yeast phenotype stimulated by FA may be due to the disruption of an existing protein interaction, by changing the stability of the protein (or), by modulating the ability of the protein to interact with other molecules, (or) by initiating a series of signal transduction pathways upon binding to a particular protein. Following the complexity of protein-ligand interactions and fully characterizing by experimental means its effects on the protein-protein interactions is challenging.
To extend our findings to human cells and identify proteins that could serve as drug targets, we replaced the yeast FMP43 protein with its human ortholog BRP44 in the genetic background of the yeast strain Δfmp43. The conservation of the two proteins was phenotypically evident, with BRP44 restoring the normal specific growth rate of the wild type, which was significantly increased by FMP43 deletion. PPIs have been proven crucial for all biological processes. Hence, by performing PPI studies it is feasible to assign functions to uncharacterized proteins and understand the composition of protein complexes.
Taking into account the high potential of human PPIs for understanding disease mechanisms and signaling cascades, we investigated a representative part of the recently described human interactome 
. We identified three interaction partners for the BRP44 protein, i.e. MAGED1, GABARAP and ACTC, all being disease-associated proteins according to the OMIM morbidmap (NCBI). Expression of members of several tumor-associated antigen families, as for instance of the MAGE family, is restricted to tumor cells and testes 
. GABARAP is a GABA-A receptor-associated protein. Type-A receptors for the neurotransmitter GABA are ligand-gated chloride channels that mediate inhibitory neurotransmission. GABARAP expression has been detected in all tissues tested, namely heart, brain, placenta, lung, liver, skeletal muscle, kidney and pancreas, suggesting potential involvement of this protein in biological events other than interaction with GABA-A receptors 
. Morgensen et al. 
stated that ACTC1
was the first sarcomeric gene described to cause two different cardiomyopathies when being mutated, and hypothesized that ACTC1
mutations affecting sarcomere to the surrounding syncytium lead to dilated cardiomyopathy. In addition, the Human Protein Reference database indicates one more very interesting interaction of BRP44 with the ribosomal P1 protein. In the recent study of Martinez-Azorin et al. 
, the role of the ribosomal stalk P proteins modulating ribosomal activity was investigated in human cells using RNA interference. The loss of P1 protein produced a decrease in the growth rate of the cells −although the details of this association are not yet understood− which is in agreement with the growth effects observed in our study when BRP44, an interaction partner of P1 protein, was functionally expressed in yeast cells.
Molecular docking methods have been used for decades to determine the affinity of ligands/substrates towards the receptor/protein/enzyme, and thus have acquired a great importance in modern structure-based drug design 
. A basic prerequisite for docking studies is the 3D structure of the proteins under study. Many proteins targeted for drug design do not have an experimentally determined structure, which makes the scope of docking studies limited. Homology modeling has been used to generate structural models of proteins, which can still be used as docking targets 
. We predicted the 3D structure of the yeast FMP43 protein, which is transcriptionally regulated after FA addition. Homology modeling was also applied to predict the structure of the human BRP44 protein, which restored the normal phenotype in the FMP43 deletion yeast strain.
The second prerequisite for docking studies is to know the location of the ligand's binding site. The information related to binding sites can be obtained experimentally through co-crystallization of the protein-ligand complex. An alternative approach is to identify structural or sequence similarity with a known binding site, or use a computational tool to predict binding sites on the protein of interest 
. The binding sites in homology models of the FMP43 and BRP44 proteins were computationally predicted, and further docking studies were performed using FA as the ligand. The results showed the affinity of FA towards both proteins, which is in line with our phenotypic experimental observations.
Comparing the protein-protein and protein-ligand interactions of simple cellular systems, for instance the yeast with the human system, promises much for the future −even though it limits the resolution of the results. The platform presented here strongly suggested anti-oxidant therapeutic targets −as demonstrated through identification and characterization of yeast and human orthrolog protein-protein interaction networks− but requires further in vivo validation. Similar to pharmacological therapeutic discovery, nutraceutical target identification and screening −while often limiting and resource-intensive− may be enhanced through the approach of yeast physiological and network biology analysis demonstrated here.