Search for: All records

Abstract Background & Aims Cancer metastasis into distant organs is an evolutionarily selective process. A better understanding of the driving forces endowing proliferative plasticity of tumor seeds in distant soils is required to develop and adapt better treatment systems for this lethal stage of the disease. To this end, we aimed to utilize transcript expression profiling features to predict the site-specific metastases of primary tumors and second, to identify the determinants of tissue specific progression. Methods We used statistical machine learning for transcript feature selection to optimize classification and built tree-based classifiers to predict tissue specific sites of metastatic progression. Results We developed a novel machine learning architecture that analyzes 33 types of RNA transcriptome profiles from The Cancer Genome Atlas (TCGA) database. Our classifier identifies the tumor type, derives synthetic instances of primary tumors metastasizing to distant organs and classifies the site-specific metastases in 16 types of cancers metastasizing to 12 locations. Conclusions We have demonstrated that site specific metastatic progression is predictable using transcriptomic profiling data from primary tumors and that the overrepresented biological processes in tumors metastasizing to congruent distant loci are highly overlapping. These results indicate site-specific progression was organotropic and core features of biological signaling pathways are identifiable that may describe proliferative plasticity in distant soils. 

Understanding spatial expressions and using them appropriately is necessary for seamless and natural human-machine interaction. However, capturing the semantics and appropriate usage of spatial prepositions is notoriously difficult, because of their vagueness and polysemy. Although modern data-driven approaches are good at capturing statistical regularities in the usage, they usually require substantial sample sizes, often do not generalize well to unseen instances and, most importantly, their structure is essentially opaque to analysis, which makes diagnosing problems and understanding their reasoning process difficult. In this work, we discuss our attempt at modeling spatial senses of prepositions in English using a combination of rule-based and statistical learning approaches. Each preposition model is implemented as a tree where each node computes certain intuitive relations associated with the preposition, with the root computing the final value of the prepositional relation itself. The models operate on a set of artificial 3D “room world” environments, designed in Blender, taking the scene itself as an input. We also discuss our annotation framework used to collect human judgments employed in the model training. Both our factored models and black-box baseline models perform quite well, but the factored models will enable reasoned explanations of spatial relation judgements.