Machine Learning Algorithms to Predict Breast Cancer Recurrence Using Structured and Unstructured Sources from Electronic Health Records.

Recurrence is a critical aspect of breast cancer (BC) that is inexorably tied to mortality. Reuse of healthcare data through Machine Learning (ML) algorithms offers great opportunities to improve the stratification of patients at risk of cancer recurrence. We hypothesized that combining features from structured and unstructured sources would provide better prediction results for 5-year cancer recurrence than either source alone. We collected and preprocessed clinical data from a cohort of BC patients, resulting in 823 valid subjects for analysis. We derived three sets of features: structured information, features from free text, and a combination of both. We evaluated the performance of five ML algorithms to predict 5-year cancer recurrence and selected the best-performing to test our hypothesis. The XGB (eXtreme Gradient Boosting) model yielded the best performance among the five evaluated algorithms, with precision = 0.900, recall = 0.907, F1-score = 0.897, and area under the receiver operating characteristic AUROC = 0.807. The best prediction results were achieved with the structured dataset, followed by the unstructured dataset, while the combined dataset achieved the poorest performance. ML algorithms for BC recurrence prediction are valuable tools to improve patient risk stratification, help with post-cancer monitoring, and plan more effective follow-up. Structured data provides the best results when fed to ML algorithms. However, an approach based on natural language processing offers comparable results while potentially requiring less mapping effort.

Interoperable Medical Application for CTC Counting.

In the present poster we will explain how the development of an interoperable AI-powered application for Circulating Tumor Cells (CTCs) counting is addressed. We will explain the selection of the most appropriate information for early detection of distant metastasis, local recurrence and the data structure definition to be compliant with international standards and ontologies.

CASIDE: A data model for interoperable cancer survivorship information based on FHIR.

Cancer survivorship has traditionally received little research attention although it is associated with a variety of long-term consequences and also many other comorbidities. There is an urgent need to increase research on this area, and the secondary use of healthcare data has the potential to provide valuable insights on survivors' health trajectories. However, cancer survivors' data is often stored in silos and collected inconsistently. In this study we present CASIDE, an interoperable data model for cancer survivorship information that aims to accelerate the secondary use of healthcare data and data sharing across institutions. It is designed to provide a holistic view of the cancer survivor, taking into account not just the clinical data but also the patient's own perspective, and is built upon the emerging Health Level Seven (HL7) Fast Healthcare Interoperability Resources (FHIR) standard. Advantages of adopting FHIR and challenges in information modelling using this standard are discussed. CASIDE is a generalizable approach that is already being used as a support tool for the development of downstream applications to support clinical decision making and can contribute to translational collaborative research on cancer survivorship.

PattRec: An easy-to-use CNV detection tool optimized for targeted NGS assays with diagnostic purposes.

Genetic laboratories use custom-commercial targeted next-generation sequencing (tg-NGS) assays to identify disease-causing variants. Although the high coverage achieved with these tests allows for the detection of copy number variants (CNVs), which account for an important proportion of the genetic burden in human diseases, an easy-to-use tool for automatic CNV detection is still lacking. This article presents a new CNV detection tool optimized for tg-NGS data: PattRec. PattRec was evaluated using a wide range of data, and its performance compared with those of other CNV detection tools. The software includes features for selecting optimal controls, discarding polymorphic CNVs prior to analysis, and filtering out deletions based on SNV zygosity, and automatically creates an in-house CNV database. There is no need for high level bioinformatic expertise and users can choose color-coded xlsx output that helps to prioritize potentially pathogenic CNVs. PattRec is presented as a Java based GUI, freely available online: https://github.com/irotero/PattRec.

Free-access copy-number variant detection tools for targeted next-generation sequencing data.

Copy number variants (CNVs) are intermediate-scale structural variants containing copy number changes involving DNA fragments of between 1 kb and 5 Mb. Although known to account for a significant proportion of the genetic burden in human disease, the role of CNVs (especially small CNVs) is often underestimated, as they are undetectable by traditional Sanger sequencing. Since the development of next-generation sequencing (NGS) technologies, several research groups have compared depth of coverage (DoC) patterns between samples, an approach that may facilitate effective CNV detection. Most CNV detection tools based on DoC comparisons are designed to work with whole-genome sequencing (WGS) or whole-exome sequencing (WES) data. However, few methods developed to date are designed for custom/commercial targeted NGS (tg-NGS) panels, the assays most commonly used for diagnostic purposes. Moreover, the development and evaluation of these tools is hindered by (i) the scarcity of thoroughly annotated data containing CNVs and (ii) a dearth of simulation tools for WES and tg-NGS that mimic the errors and biases encountered in these data. Here, we review DoC-based CNV detection methods described in the current literature, assess their performance with simulated tg-NGS data, and discuss their strengths and weaknesses when integrated into the daily laboratory workflow. Our findings suggest that the best methods for CNV detection in tg-NGS panels are DECoN, ExomeDepth, and ExomeCNV. Regardless of the method used, there is a need to make these programs more user-friendly to enable their use by diagnostic laboratory staff who lack bioinformatics training.

Linked Registries: Connecting Rare Diseases Patient Registries through a Semantic Web Layer.

Patient registries are an essential tool to increase current knowledge regarding rare diseases. Understanding these data is a vital step to improve patient treatments and to create the most adequate tools for personalized medicine. However, the growing number of disease-specific patient registries brings also new technical challenges. Usually, these systems are developed as closed data silos, with independent formats and models, lacking comprehensive mechanisms to enable data sharing. To tackle these challenges, we developed a Semantic Web based solution that allows connecting distributed and heterogeneous registries, enabling the federation of knowledge between multiple independent environments. This semantic layer creates a holistic view over a set of anonymised registries, supporting semantic data representation, integrated access, and querying. The implemented system gave us the opportunity to answer challenging questions across disperse rare disease patient registries. The interconnection between those registries using Semantic Web technologies benefits our final solution in a way that we can query single or multiple instances according to our needs. The outcome is a unique semantic layer, connecting miscellaneous registries and delivering a lightweight holistic perspective over the wealth of knowledge stemming from linked rare disease patient registries.

SCALEUS: Semantic Web Services Integration for Biomedical Applications.

In recent years, we have witnessed an explosion of biological data resulting largely from the demands of life science research. The vast majority of these data are freely available via diverse bioinformatics platforms, including relational databases and conventional keyword search applications. This type of approach has achieved great results in the last few years, but proved to be unfeasible when information needs to be combined or shared among different and scattered sources. During recent years, many of these data distribution challenges have been solved with the adoption of semantic web. Despite the evident benefits of this technology, its adoption introduced new challenges related with the migration process, from existent systems to the semantic level. To facilitate this transition, we have developed Scaleus, a semantic web migration tool that can be deployed on top of traditional systems in order to bring knowledge, inference rules, and query federation to the existent data. Targeted at the biomedical domain, this web-based platform offers, in a single package, straightforward data integration and semantic web services that help developers and researchers in the creation process of new semantically enhanced information systems. SCALEUS is available as open source at http://bioinformatics-ua.github.io/scaleus/ .