Towards in silico-guided clinical trials in cancer
Our Subject Matter Expert Anne Lise Børresen Dale – internationally famous in the field of Breast Cancer – invited PubGene to a workshop organized by the faculty of Medicine of Oslo, bringing international experts in fields such as systems medicine, mathematical oncology and bioinformatics to discuss novel clinical trial concepts for personalized cancer medicine.
Here are just a few of the highlights:
In silico vs. clinical tumor board, P. Fasching
Peter Fasching, MD, took us through an entertaining journey where human and machine worked hand in hand in order to augment the brain capacity of physicians on Breast Cancer therapy decision. The algorithm was fed with historical patient data and the effective therapy decisions, and it learned to approximate the physician’s therapeutic choice on new patient data. Interestingly, in a blinded study, the therapy choices proposed by this Machine learning based Clinical Decision Support developed in the context of a research project (PRAEGNANT), were more trusted than those of tumor boards second opinions from supposed colleagues.
The PRAEGNANT study is a real-time, real-world registry for metastatic breast cancer patients. Real-world data may help to clarify whether the efficacy of therapy sequences is as least as good as the approval study.
Peter’s talk ended with some open questions: “How do we bring patients, doctors, and scientists together to use new genomic tools and machine learning?”, “How do we give you access to the data?” Interesting! At PubGene, our Coremine Vitae team strives to answer those same questions. Our researchers and Subject Matter Experts use the technology developed by bioinformaticians to empower people with information about their genome and phenome, so that their doctor can find the most optimal treatment plan. We are working on a digital health solution to do it at scale. Our biologist Håvard Hauge had a good talk with Peter over lunch and concluded that a plane ticket to Germany does not cost much. Could Coremine Vitae’s breast cancer cases benefit from this encounter?
Therapeutics targets from Big Data, R. Jörnsten
Rebecka Jörnsten presented the work of her colleagues at the Sven Nelander Lab, related to the network modeling of cancer and data integration for discovery of treatments, focusing on children with high-risk neuroblastoma.
As part of the research, the group developed a new algorithm, available on http://targettranslator.org/ , which combines data from tumor biobanks, pharmacological databases, and cellular networks, to predict how particular targeted interventions will affect mRNA signatures associated with high patient risk.
In this public tool, you can input clinical variables (e.g. age, survival time, mutations) and log2 transformed expression values. As you select risk factors from your clinical or gene variables, they get highlighted in the gene signature heatmap, with high values more colored. The result is a list of drugs most likely to induce the change you wanted to investigate, and the protein targets as defined by the STITCH database that seems to be enriched in the drug scoring table.
The group expects this method to accelerate the discovery of risk-associated targets for cancers. Looking forward to discussing this with my colleagues at Coremine Vitae.