Biological Models Volume Title: From Knowledge Networks to Biological Models

Androgen deprivation therapy (ADT) is the gold standard for advanced stages of prostate cancer due to the good patient responding to this treatment. Nevertheless, in most cases the disease acquires hormone-refractory status which is considered incurable. The aim of this study is to draw a mechanistic model of androgen receptor (AR) signaling in normal prostate and cancerous tissue in order to better understand the causes of treatment failure. With assistance of Pathway Studio software (Ariadne Genomics Inc., USA) we have created overview pathway showing principal signaling cascades disturbed during prostate cancer development and progression. We further discuss the role of androgen receptor in homeostasis of prostate normal cells and in deregulating signaling of cancer cells. Due to AR involvement not only in proliferation, but also in apoptosis, invasive growth and in cell-to-cell communication, we hypothesized that ADT in some cases may contribute to the development of the castration resistant cancer. We review that upon androgen deprived conditions impaired AR activity can lead to the disturbances in multiple pathways thereby influencing global homeostasis of the prostate. We concluded that potential of AR to potentiate prostate cancer rather than inhibiting it should be taken into consideration when choosing ADT as an option for prostate cancer management.

Familial amyotrophic lateral sclerosis (fALS) is a hereditary disorder of motor neurons that is caused by mutation in Cu, Zn superoxide dismutase (SOD1) in a subset of cases. The onset of the disease is relatively late, usually at age 50 or later, and is associated with interrelated molecular mechanisms of neurodegeneration. One of the mechanisms that can promote ALS progression is increased intracellular calcium concentration. The only market-available drug for ALS targets glutamate receptors and slows disease in part by mitigating excitotoxicity, a process in which persistent stimulation of glutamate receptors leads to pathologically high calcium concentration. To dissect the potential contributions of calcium mishandling to ALS, we have processed several publically available expression datasets related to fALS and analyzed the differential expression of genes related to calcium homeostasis. We find that SOD1- related fALS is associated with changes in expression of numerous genes related to calcium handling. Several genes which are down-regulated in fALS are targets of the repressor element-1 transcription factor/neuron restrictive silencer factor (REST/NRSF) transcription factor, which is normally inactivated in neuronal tissue. Our meta-analysis shows that changes in gene expression occurring in SOD1-related fALS promote calcium mishandling through dysregulation of multiple pathways, and that aberrant REST/NRSF activity may underlie some errors in calcium homeostasis.

We describe construction of mechanistic model for drug-induced cholestasis using information available from ResNet and ChemEffect knowledge networks in Pathway Studio software. We first developed the mechanistic model using information about protein targets of the cholestasis-inducing drugs. We then expanded the model by incorporating knowledge about protein functional annotation, protein homology, canonical pathways and pathway reconstruction. The expanded model provides a toxicity mechanism for 81% of the drugs known to induce cholestasis vs. 58% of the drugs explained by the original model. Using the model we suggest that FGF19 secreted proteins are the biomarker for drug-induced cholestasis. In this discussion we suggest how the mechanistic model can be used for predicting cholestasic risk for new compounds, how it can be used for development of biomarker panel to monitor cholestasis risk in patient during drug therapy, and how the model can be used in personalized medicine for evaluating patient predisposition for cholestasis.

The underlying cause of Duchenne muscular dystrophy (DMD) – mutations in the dystrophin gene – is known for 25 years. Still many details are to be elucidated to reconstruct the complete picture of DMD pathogenesis explaining how the lack of dystrophin leads to the disease symptoms. Dissecting the complex disease into a set of disturbed pathways helps organizing already known facts and discovering new nodes important for disease progression.

We suggest three approaches to characterize DMD through pathways. First, we manually built DMD pathways based on the literature evidence to show how intersecting disease-specific pathways allows identification of common regulators in DMD which might be considered as potential drug targets. Second, we used algorithmically generated subnetworks and a set of curated expression targets pathways to analyze genes that change expression in DMD. Using collection of the predefined pathways or automatically generated subnetworks for data analysis reveals new nodes (e.g. ESRRA and SREBF1) and pathways (e.g. IL6 and IGF1 signaling) crucial for the disease but not yet covered in literature.

We have built a model for gastric cancer predisposition caused by hypergastrenimia during Helicobacter infection. The model was built using publically available data form hypergastrenimic transgenic mice infected with Helicobacter. We used the model to identify potential drug targets for gastric cancer, annexin II and TRAF6, and to find prognostic biomarkers for gastric cancer predisposition due to hypergastrenimia.

Patient stratification or, a personalized approach to medical treatment, is a promising approach in modern medicine. Finding biological patterns within a group of patients with the same diagnosis could lead to more precise and effective therapies. To address this issue it is necessary to reveal different mechanisms within the same disease, to find new biomarkers, and to develop new diagnostic tests that would distinguish patients from different subgroups.

Cancer sub-typing based on clustering of individual patient gene expression profiles has been widely used for various types of cancer. Here we propose a new approach which includes the consecutive use of Sub-network enrichment analysis algorithm (SNEA) for individual differential expression profiles and biclustering of found expression regulators and samples.

We analyzed nine publicly available microarray datasets with data from patients suffering from colorectal cancer as compared to healthy donors, including one dataset containing supplementary information on patient response to anti-EGFR therapy with cetuximab. We have identified several patient subtypes characterized by specific regulatory clusters (pathways) and mapped the data about cetuximab response onto the heat map of pathway activity for each patient. We found that the most prominent mechanism that distinguished responders from non-responders is dependent on regulators from the TGF-β/SMAD pathway and corresponds to the epithelial-tomesenchymal transition (EMT).