BioVariance is a member of the EU-funded REVERT Project

Colorectal cancer is one of the most common types of cancer worldwide. People suffering from metastatic colorectal cancer (mCRC) often have a bad prognosis of survival. Only 20 % of patients survive longer than 5 years after being diagnosed with mCRC. Additionally, commonly used therapies do not work for all patients equally well. Therefore, a lot of research is done in the field of colorectal cancer. Due to those reasons BioVariance participates in the REVERT project with the aim of improving and individualizing the treatment of mCRC.


What is metastatic colorectal cancer?
The third most common type of cancer is colorectal cancer. mCRC is the metastatic type of colorectal cancer being located at the colon or rectal part of the human body. It appears more often in men than in women, but women often suffer from the more aggressive right-sided type of the cancer. The right-sided type of mCRC goes along with a worse prognosis of survival. Generally, one can distinguish between left- and right-sided colorectal cancer, where left-sided mCRC patients have better prognoses concerning the progression-free survival (PFS). The PFS is a factor that measures how long people survive after the cancer treatment without the tumor progressing again. A lot of mCRC cases are related to higher ages, because age is one of the risk factors. Other risk factors are a bad lifestyle (bad diet habits, alcohol consumption, etc) and less frequently also genetic factors.


Classification of mCRC patients is done by the oncologists on basis of anatomical tumor characteristics like grading and staging. Grading indicates how similar a tumor cell is to a normal cell, whereas staging gives information about the size and the spreading of the tumor. Since not all patients have the same present biomarkers (biologically measurable parameters of biological processes, that are known to have prognostic or diagnostic expressive power) and clinical data (measured values like weight, height, blood pressure, vitamin d value etc.), not all treatments will work equally well for all patients. Thus, a more personalized approach is being aspired, where in addition to grading and staging several other molecular biomarkers and individual clinical data of the patient play a more important role in the treatment decision.


What does REVERT mean?
The abbreviation REVERT origins from the long title “taRgeted thErapy for adVanced colorEctal canceR paTients”. REVERT is an international project funded by the EU, that started in 2020 and is planned to end in 2023. Specialists from Italy, Spain, Sweden, Luxembourg, Romania, and Germany were chosen to participate in the project, involving clinicians, pathologists, biologists, software developers, and also artificial intelligence (AI) specialists. BioVariance is part of the work package being responsible for the development of the AI solutions, thus BioVariance will offer some machine learning solutions for the question of the project.


What is the aim of the project?
The project aims at generating the REVERT database of molecular and clinical data collected from several hundred metastatic colorectal cancer patients that had visited one of the participating hospitals located at the different countries within the EU in the past. The database is then used as a basis for machine learning models to aspire a more personalized therapy approach for treating newly diagnosed patients. In the future, the trained AI model will predict a suitable first-line treatment for each new patient, who has not received any cancer medication yet, supporting the clinicians in their treatment decision. In addition to that, a clinical study will be executed with several new patients that decide to participate and want to be treated according to the AI-based results. This prospective clinical study will be performed with already authorized and commonly used chemotherapies and not with new treatments. The expectation concerning the AI-based results is to provide another more objective decision factor for clinicians, that is free from the subjective decision of the attending doctor at the hospital. Finally, the machine learning (ML) solutions will be combined in an app, that can support clinicians in the future.


What is machine learning and how is it used in the context of the project?
Machine learning algorithms use underlying mathematical models to recognize certain patterns in the patient’s data. The primary dataset of the REVERT database contains several processes of cancer treatments, the known survival outcome and about 150 other molecular and clinical features of hundreds of mCRC patients from the last 10 years. According to the underlying mathematical models, it is possible to derive some new rules from this dataset, that make it possible to predict outcomes for each new patient that was not included in the primary data. Therefore, the primary dataset must be used to (i) train and thereby generating ML models and to (ii) test and evaluate these generated models. In addition, the ML models allow to spot important features, that have a strong influence on the outcome prediction. For the primary dataset the actual outcome is already known, allowing us to see for how many patients the model predicted the correct outcome.


In the REVERT project BioVariance’s ML algorithms are used to create models that can classify whether a certain patient is likely to respond to different chemotherapies, that have been approved for years. During the project, several chemotherapies have already been categorized by the clinicians as being equally well for treating mCRC and that are now considered for the project. The related question for the ML algorithms is to predict the first-line treatment (= initial treatment, that is recommended for a disease, here: chemotherapy) having the highest chance of success for a certain patient.


What is the current status of the project and what is still to come?
At the moment, the machine learning models are already trained and tested on the previously collected dataset. BioVariance’s models and the models of three other REVERT members are used to predict the best first-line treatment for real patients participating in a phase two clinical study that started in March this year. The first few patients that were diagnosed with mCRC and enrolled to the study have already been processed with the ML models and were treated according to the model outcomes.


The main focus lays on the continuous improvement of the ML algorithms to get the best predictions possible. In the near future the ML models will be included into the REVERT app interface. This allows clinicians to directly enter patient’s data into the REVERT app and get a prediction for the most suitable first-line treatment on a patient-based level.


This approach is an important step into the direction of personalized medicine, since the huge amount of each patient’s data is intensively analyzed and leads to the best treatment for that patient individually, whereas in the current decision process patients get more or less grouped into stages or grades and are then treated in the same way. Finally, the clinical study will show whether the more personalized approach will be able to improve the survival of the patients.