Create a medical cannabis patient database
As of 2016, over 25,000 people in Israel are currently using physician-prescribed medical Cannabis. Although these patients are under medical supervision, physicians usually do not follow up on the medicinal benefits or side effects of the Cannabis treatment. Patients who are prescribed Cannabis are typically directed to a Cannabis grower or distributer and it is the grower who essentially decides which strain of Cannabis to provide and the route of administration. Unfortunately, the latter information often remains with the Cannabis provider and is not brought back to the attention of the physician who prescribed the treatment.
This unique situation leaves the care of the patients with the Cannabis providers who are not qualified health care providers. In addition, this course of treatment creates a situation in which the data is divided between the physicians in the hospital, the community caregivers, and the growers. Each one of the aforementioned groups collects data related to his part of the treatment without integrating it with the knowledge of the others and in most cases, without any comparative work which can give prospective knowledge.
We aim to construct an integrated database with both clinical data on Cannabis patients as well as data on the Cannabis usage history for Cannabis patients in Israel and abroad. Acquired data from growers, Health Maintenance Organizations (HMOs), and other sources will be used for initial statistical and computational analyses, aimed to identify patient-class specific treatments and organize them into an accessible website for patients and physicians. Additionally, we aim to develop an automated database synchronization method to update our patient database from the data maintained by the HMOs and Cannabis growers.
Cannabis strain database
In addition to the patient database, we are creating a Cannabis strain database. Using state-of-the-art mass-spectrometry, we are comprehensively profiling the cannabinoid composition for a variety of Cannabis strains that are used for clinical purposes in Israel. We have already profiled more than 80 strains and expect to have all the Israeli Cannabis strains uploaded to the stain database up and running before the end of the 2017. The patient-specific and Cannabis strain databases are combined, analyzed, and used to derive optimized treatment options thereby maximizing disease specific therapeutic efficacy while minimizing side-effects.
Our first goal is to assign each patient class (disease, age, gender, medical data, etc.) with the most effective Cannabis treatment (strain, dose, and route of administration). For each patient class, we further analyze the risk of side effects due to usage of different Cannabis treatments against the patients’ genetic backgrounds. This provides means for evidence-based recommendations for the most suitable, safe and effective Cannabis treatment.
Integrating these various datasets and being able to match disease classes with optimized Cannabis treatments is challenging, considering the complex nature of the obtained datasets. To address this, we apply computational clustering and machine learning techniques that are commonly applied in the field of bioinformatics. Specifically, we evaluate the usage of two computational classification techniques that are commonly used in prediction of drug efficacy versus side effects: Bayesian networks and Support Vector Machines (SVM). The computational model is optimized via cross-validation tests. For each patient class the model produces: i) a ranked list of optimal Cannabis treatments, along with ii) the estimated probability of having specific therapeutic efficacy and iii) the risk of specific side effects.