AI Services

Baseline

Prof. MD. Michel Aguet of Molecular Biology at the EPF Lausanne wanted to find out whether the Exploris AI platform could be used to make the optimal therapy choice for individual breast cancer patients. Identifying the most effective, individually tailored therapy is a long-sought goal in breast cancer treatment. Reliable prediction of breast cancer progression and its survival is key to individualized therapy. Most patients receive more therapies than necessary. For some of these patients, surgery might be enough, others may need additional treatments such as chemotherapy, and still, others may need additional hormone therapy, radiation, or more.

Result

The AI-X-platform was able to successfully distinguish between high and low activity. This means recognizing whether the tumor is growing stronger or weaker, which is essential for selecting a therapy. The Dutch model from the Van de Vijver (2002) study was used as a basis for comparison. The evaluation of the results of the model using the Kaplan-Meier curve outperformed the results which can be seen from the graph. In comparison our model shows a 30 percent improvement in prediction of breast cancer progression which results in improved therapy selection.

‍

Baseline

Finding the right active ingredient faster while saving 90% of the usual animal testing (pre-study) can be achieved with artificial intelligence. Every approved drug is preceded by a long search for active ingredients. Traditionally, thousands of molecules are screened to identify the most promising candidates. A leading pharmaceutical company in allergy research wanted to find the most promising candidate with the highest treatment effect from two drug candidates for mite allergy. The study compared the two compounds against a placebo.

Result

We were able to separate the placebo effect from the treatment effect and presented that stratification can identify future candidates. Thus, the potentially most promising candidate was identified.

Baseline

A pharmaceutical company wanted to optimize the efficiency of a phase 3 trial for the development of an Alzheimer's drug. The main problem of the trial was that many patients showed a deterioration of cognitive performance despite drug treatment. In general, the inclusion of relevant patients in clinical trials is crucial to their success. Especially in diseases with unclear or ambiguous symptoms, it is important to ensure that those whose symptoms are attributable to the disease under study are included in the study.

Result

Data analysis using the AI-X-platform revealed that patients who did not have Alzheimer's disease were included in the study. Therefore, the drug could not achieve the desired effect. Correct stratification of patients will be used in the future to demonstrate a higher efficacy of the drug.

Baseline

New biomarkers should allow improved diagnosis and therapy of certain diseases. For Boehringer, we looked for biomarkers that allow the differentiation between two diseases (Lupus Erythematosus and Arthritis). This involved analyzing 24,000 proteins per patient (proteomics data) and using artificial intelligence to find the relevant biomarkers.

Result

36 relevant gene sets were found, which allow clear differentiation of patients between Lupus Erythematosus and Arthritis. In addition, we discovered that these found genes are responsible for the regulation of cell death.