# Context Interventional cancer clinical trials are generally too restrictive, and some patients are often excluded on the basis of comorbidity, past or concomitant treatments, or the fact that they are over a certain age. The efficacy and safety of new treatments for patients with these characteristics are, therefore, not defined. In this work, we built a model to automatically predict whether short clinical statements were considered inclusion or exclusion criteria. We used protocols from cancer clinical trials that were available in public registries from the last 18 years to train word-embeddings, and we constructed a dataset of 6M short free-texts labeled as eligible or not eligible. A text classifier was trained using deep neural networks, with pre-trained word-embeddings as inputs, to predict whether or not short free-text statements describing clinical information were considered eligible. We additionally analyzed the semantic reasoning of the word-embedding representations obtained and were able to identify equivalent treatments for a type of tumor analogous with the drugs used to treat other tumors. We show that representation learning using deep neural networks can be successfully leveraged to extract the medical knowledge from clinical trial protocols for potentially assisting practitioners when prescribing treatments. # Content A total of 6,186,572 labeled clinical statements were extracted from 49,201 interventional CT protocols on cancer (the URL for downloading this dataset is freely available at https://clinicaltrials.gov/ct2/results?term=neoplasmtype=Intrshowdow). Each CT downloaded is an XML file that follows a structure of fields defined by an XML schema of clinical trials [16]. The relevant data for this project are derived from the intervention, condition, and eligibility fields written in unstructured free-text language. The information in the eligibility criteria—both exclusion and inclusion criteria—are sets of phrases and/or sentences displayed in a free format, such as paragraphs, bulleted lists, enumeration lists, etc. None of these fields use common standards, nor do they enforce the use of standardized terms from medical dictionaries and ontologies. Moreover, the language had the problems of both polysemy and synonymy. The original data were exploited by merging eligibility criteria together with the study condition and intervention, and subsequently transforming them into lists of short labeled clinical statements that consisted of two extracted features (see example in Figure 2), the label (Eligible or Not Eligible), and the processed text that included the original eligibility criterion merged with the study interventions and the study conditions. See more details: Bustos, A.; Pertusa, A. Learning Eligibility in Cancer Clinical Trials Using Deep Neural Networks. Appl. Sci. 2018, 8, 1206. www.mdpi.com/2076-3417/8/7/1206 [www.mdpi.com/2076-3417/8/7/1206][1] [1]: http://www.mdpi.com/2076-3417/8/7/1206