Precision Medicine and Artificial Intelligence with SAS Vaya

SAS Viya Healthcare Life Science


In this article, we will analyze some possible applications of artificial intelligence in Healthcare and Life Science, with the aim of promoting research development and improving clinical approaches towards patients. The focus will be on data analysis utilizing the SAS Viya platform, specifically the SAS Visual Analytics module.

What is SAS Visual Analytics?

SAS Visual Analytics is the data visualization tool of SAS Viya, allowing for quick analysis and visual exploration of large datasets and creation of interactive dashboards. In addition to classic data visualization, the tool includes advanced analytics modules, providing visual objects for implementing machine learning algorithms.

One of the most recent and challenging endeavors in healthcare is finding the best treatment for individual patients. It is well-known that there is variability in treatment response in terms of efficacy/risks among different patients due to various factors such as sociodemographic characteristics, genetic variability, lifestyle, and environment: this is referred to as personalized or precision medicine.

Precision medicine can be further categorized into:

  • Precision prevention: using data related to biological, behavioral, socioeconomic, and epidemiological parameters to outline and implement tailored strategies for individuals or communities;
  • Precision diagnostics: diagnosing diseases based on individual omics data;
  • Precision treatment: enhancing outcomes through targeted and personalized treatments.

Precision prevention

Starting with precision prevention, there are documented cases in literature showcasing potential applications of artificial intelligence. Particularly, by constructing a machine learning model, one can estimate the risk of an individual patient developing a certain condition, based on specific biomarkers derived from a sample. However, in reality, these cases often face challenges such as limited population data availability and rarity of the target of interest. Managing dataset partitioning (training, validation, and test) becomes a significant issue in such scenarios.

However, the cases reported in literature often clash with the harsh reality of limited data availability from the populations under analysis and the rarity of the target of interest. In such scenarios, managing dataset partitioning (training, validation, and test) becomes a significant challenge. The sparse observations of the target provide a limited and non-comprehensive view of the phenomenon. Additionally, when splitting into training and validation sets, new misalignments arise. This is because the partitioning may separate targets of one type in the training set from targets of another type in the validation set, thereby testing a model fitted on a target configured in a slightly or heavily different manner. Hence, contents should ideally be distributed as homogeneously as possible, although unfortunately, this is not always feasible.

SAS Visual Analytics Visual Interfaces

When approaching a new analytical case, it’s important to structure oneself to optimize time, leveraging the most efficient strategy. In order to do this, it’s crucial to quickly compare different strategies to highlight the most effective one in the shortest time possible. Thanks to the visual interfaces available in SAS Visual Analytics, it’s possible to swiftly explore data characteristics from a descriptive standpoint as well as implement machine learning models. This objective is particularly achievable when, as in the case of Visual Analytics, the visual interface is combined with in-memory processing, which is highly responsive and provides immediate feedback on the effectiveness of the strategy being tested at that moment.

SAS Model Studio for Machine Learning Model Testing

Another tool available on the SAS Viya platform is SAS Model Studio (evolution of Enterprise Miner), enabling the construction of visual analytical pipelines to test various ML models in parallel and select the most performant approach through the node of comparison model. Additionally, the pipeline facilitates comparisons between different ML models in terms of performance and methodologies for data preparation and transformation.

SAS Visual Analytics Healthcare Life Science

Precision diagnostic

Moving on to precision diagnostics, in this case as well, one can rely on the use of a specific biomarker for the diagnostic identification of a certain pathology (e.g., neurodegenerative, cardiovascular, gestational diseases): in this scenario, we are talking about diagnostics as the aim is to intercept those situations where the pathology is already present.

The data that needs to be brought together can be quite diverse: for example, a specific biomarker, a series of clinical information (not necessarily in the form of structured data, e.g., text in the case of a medical report), the patient’s clinical history evolution; this array of diverse information represents the input for analysis.

In this case too, the use of SAS Visual Analytics, in addition to operators for traditional visualization (e.g., bar chart, pie chart), allows for data analysis using visual objects that create machine learning models in drag and drop mode without the need for prior knowledge of SAS code.

Precision treatment

Regarding precision treatment, cases are still very limited. However, it’s interesting to highlight how in this field, the integration and utilization of omics data are crucial for a specific personalized therapeutic approach. Omics data refers to all data related to an individual and encompasses aspects of genomics, transcriptomics, and metabolomics. In this context, it’s important to analyze the existing correlation between specific genetic patterns and molecules of interest. Having access to a list of potential drugs used for a certain pathology and the individual’s genetic information, it’s necessary to cross-reference this data to achieve the association that guarantees the best effect with reduced adverse effects. The use of this data allows for further patient stratification based on the type of drug that best suits the specific conditions of that particular patient, thus reducing adverse reactions.

Literature also reports cases where omics data has been used to identify alterations involved in certain tumor forms, aiming to find possible correlations between the pathology and genomic or proteomic profiles using machine learning models to identify potential mechanisms involved in tumor pathogenesis.

Therefore, the application of artificial intelligence in healthcare represents an important topic that must be addressed with attention concerning the management of the increasing volume of data generated by medical research, but also with the right enthusiasm to promote responsible usage that can lead to the improvement of community health.

Blue BI and SAS Viya

At Blue BI, we have technical and scientific expertise to handle the complexity of data and business processes typical of the Life Science & Pharma sector. Our industry knowledge and extensive experience in Clinical Studies enable us to manage solutions that comply with privacy regulations and the highest security standards. Our constant commitment to innovative strategies leads us to develop advanced and cutting-edge solutions. Additionally, we have years of experience in using SAS Viya, allowing us to extract significant insights from elementary data, even from heterogeneous sources, in real-time to achieve impactful insights.

The BBI x Clinical Trials solution is a ready-to-use advanced Business Intelligence tool for efficient management of clinical studies, flexible, and scalable, capable of adapting to the needs of every client.


Table of Contents