Patient-Specific Modeling of Diffuse Large B-Cell Lymphoma.

Personalized medicine aims to tailor treatment to patients based on their individual genetic or molecular background. Especially in diseases with a large molecular heterogeneity, such as diffuse large B-cell lymphoma (DLBCL), personalized medicine has the potential to improve outcome and/or to reduce resistance towards treatment. However, integration of patient-specific information into a computational model is challenging and has not been achieved for DLBCL. Here, we developed a computational model describing signaling pathways and expression of critical germinal center markers. The model integrates the regulatory mechanism of the signaling and gene expression network and covers more than 50 components, many carrying genetic lesions common in DLBCL. Using clinical and genomic data of 164 primary DLBCL patients, we implemented mutations, structural variants and copy number alterations as perturbations in the model using the CoLoMoTo notebook. Leveraging patient-specific genotypes and simulation of the expression of marker genes in specific germinal center conditions allows us to predict the consequence of the modeled pathways for each patient. Finally, besides modeling how genetic perturbations alter physiological signaling, we also predicted for each patient model the effect of rational inhibitors, such as Ibrutinib, that are currently discussed as possible DLBCL treatments, showing patient-dependent variations in effectiveness and synergies.

PEDL: extracting protein-protein associations using deep language models and distant supervision.

MOTIVATION: A significant portion of molecular biology investigates signalling pathways and thus depends on an up-to-date and complete resource of functional protein-protein associations (PPAs) that constitute such pathways. Despite extensive curation efforts, major pathway databases are still notoriously incomplete. Relation extraction can help to gather such pathway information from biomedical publications. Current methods for extracting PPAs typically rely exclusively on rare manually labelled data which severely limits their performance. RESULTS: We propose PPA Extraction with Deep Language (PEDL), a method for predicting PPAs from text that combines deep language models and distant supervision. Due to the reliance on distant supervision, PEDL has access to an order of magnitude more training data than methods solely relying on manually labelled annotations. We introduce three different datasets for PPA prediction and evaluate PEDL for the two subtasks of predicting PPAs between two proteins, as well as identifying the text spans stating the PPA. We compared PEDL with a recently published state-of-the-art model and found that on average PEDL performs better in both tasks on all three datasets. An expert evaluation demonstrates that PEDL can be used to predict PPAs that are missing from major pathway databases and that it correctly identifies the text spans supporting the PPA. AVAILABILITY AND IMPLEMENTATION: PEDL is freely available at https://github.com/leonweber/pedl. The repository also includes scripts to generate the used datasets and to reproduce the experiments from this article. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Evaluating Uncertainty in Signaling Networks Using Logical Modeling.

Systems biology studies the structure and dynamics of biological systems using mathematical approaches. Bottom-up approaches create models from prior knowledge but usually cannot cope with uncertainty, whereas top-down approaches infer models directly from data using statistical methods but mostly neglect valuable known information from former studies. Here, we want to present a workflow that includes prior knowledge while allowing for uncertainty in the modeling process. We build not one but all possible models that arise from the uncertainty using logical modeling and subsequently filter for those models in agreement with data in a top-down manner. This approach enables us to investigate new and more complex biological research questions, however, the encoding in such a framework is often not obvious and thus not easily accessible for researcher from life sciences. To mitigate this problem, we formulate a pipeline with specific templates to address some research questions common in signaling network analysis. To illustrate the potential of this approach, we applied the pipeline to growth factor signaling processes in two renal cancer cell lines. These two cell lines originate from similar tissue, but surprisingly showed a very different behavior toward the cancer drug Sorafenib. Thus our aim was to explore differences between these cell lines regarding three sources of uncertainty in one analysis: possible targets of Sorafenib, crosstalk between involved pathways, and the effect of a mutation in mammalian target of Rapamycin (mTOR) in one of the cell lines. We were able to show that the model pools from the cell lines are disjoint, thus the discrepancies in behavior originate from differences in the cellular wiring. Also the mutation in mTOR is not affecting its activity in the pathway. The results on Sorafenib, while not fully clarifying the mechanisms involved, illustrate the potential of this analysis for generating new hypotheses.

Unraveling the regulation of mTORC2 using logical modeling.

BACKGROUND: The mammalian target of rapamycin (mTOR) is a regulator of cell proliferation, cell growth and apoptosis working through two distinct complexes: mTORC1 and mTORC2. Although much is known about the activation and inactivation of mTORC1, the processes controlling mTORC2 remain poorly characterized. Experimental and modeling studies have attempted to explain the regulation of mTORC2 but have yielded several conflicting hypotheses. More specifically, the Phosphoinositide 3-kinase (PI3K) pathway was shown to be involved in this process, but the identity of the kinase interacting with and regulating mTORC2 remains to be determined (Cybulski and Hall, Trends Biochem Sci 34:620-7, 2009). METHOD: We performed a literature search and identified 5 published hypotheses describing mTORC2 regulation. Based on these hypotheses, we built logical models, not only for each single hypothesis but also for all combinations and possible mechanisms among them. Based on data provided by the original studies, a systematic analysis of all models was performed. RESULTS: We were able to find models that account for experimental observations from every original study, but do not require all 5 hypotheses to be implemented. Surprisingly, all hypotheses were in agreement with all tested data gathered from the different studies and PI3K was identified as an essential regulator of mTORC2. CONCLUSION: The results and additional data suggest that more than one regulator is necessary to explain the behavior of mTORC2. Finally, this study proposes a new experiment to validate mTORC1 as second essential regulator.

Data-driven optimizations for model checking of multi-valued regulatory networks.

The model checking method has been long since established as an important tool for modelling and reverse engineering of biological systems. However, due to a high complexity of both the method and the biological systems, this approach often requires a vast amount of computational resources. In this article we show that by reducing the expressivity of the method we can gain performance while still being able to use all biologically relevant data. We utilize this approach to conduct a study of mutations in the EGFR signalling, motivated by a paper from Klinger et al. (2013). Here we aim at constructing approximated models of multiple cell-lines from sizeable sets of experimental data. Due to cancerous mutations in each cell line, there is a high degree of parameter uncertainty and the study would not be practically tractable without the performance optimizations described here.