By Prof. Gerd Binnig, Chief Technology Officer, Definiens
Seeing is believing – this is an important aspect of Tissue Phenomics, where cellular structures in tissue slides taken from patients are investigated in real space. Observing directly what type of structures form and what kind of interactions take place is in general of much higher value than concluding indirectly from other type of data. I experienced this factor very clearly thirty years ago when people speculated about the atomic structure on the surfaces of matter from indirect measurement. When we could, however, image atomic structures directly, it became obvious that most of those speculations were wrong. This is the reason why even for a genetic disease like cancer, genetic information could not push histology aside. Observing the structures and the interactions in tissue slides on the cell level enables insights into the disease of a specific person that is decisive and cannot be created otherwise. Tissue Phenomics builds upon histological information and pathologists’ knowledge and is by no means about replacing it. The mission is rather to gain an even deeper insight into biological processes detectable in cell structures for the enablement of new and successful treatments of patients and their specific diseases. In the beginning there was the question how far one can get with the Tissue Phenomics approach, or more concretely, how far can we go beyond what the experts already know today? What can be achieved when the computer investigates all kind of cell patterns in tissue and correlate them with biological and medical relevance and meaning, represented e.g. by clinical outcome? Today we know better.
This technology, called Tissue Phenomics, makes use of image analysis and data mining and is on the way to a game-changing big data approach. In the last few years, several groups have demonstrated that the potential of Tissue Phenomics is huge and that it might change the way patients are treated, particularly how they are selected for specific treatments. One example is the pioneering work of Jerome Galon et al. on the Immunoscore®. Another example is the work of Peter Caie et al. who applies the concept of Tissue Phenomics to both clinical tissue and cytology based projects.
There are several reasons why Tissue Phenomics is important and why it will play a central role in the future of medicine.
Let us address those five key points in more detail.
Information is moving to the center of medical care
In general, information and the extensive collection of it becomes increasingly more important and may even play the most important role in the future of medicine. Today a vast amount of information gets lost – particularly in medicine. This is because most data are simply not available. Data are mostly tacit and not explicit either because they are not digital or not documented at all. This becomes most obvious in histology where cancer patients are diagnosed in a report by very brief written statements of a pathologist. The various forms of intermediate conclusions and considerations only take place in the mind of the pathologist and are not documented. Only in very rare cases are objects like specific cells counted and statistical values derived and documented. Compared to all of the measurements that are possible and would be meaningful, this data is very, very little. There is a process ongoing that changes this situation. Medical information gets more and more explicit through digitization and by structuring data meaningfully. This way data can be investigated, sorted and used to create novel insights by statistical machine learning methods.
This process has already started in medicine, and Tissue Phenomics is becoming a crucial part of it. It will have a transformative impact on how patients will be treated in the future – i.e. much more information-driven than today.
Cellular information is key for optimal patient treatment
Although cancer is viewed as a genetic disease, from a different angle, it can be seen as an immune disease. In principle, the immune system fights cells that are non-self. The MHC complex of a cancer cell itself unfortunately is “self”, but there are many antigens presented by the tumor cell that should be recognized by the immune cells as non-self. Basically, it should not matter what kind of tumor antigens are present, i.e. what type of mutations are actually present, unless the antigens are recognized as foreign. There are drugs on the market that just stimulate the activity of the immune cells, and when administered, the tumor disappears after a while. This mechanism, however, works only for a small percentage of the cancer patients. The situation becomes even more complex when the tumor, in the course of its evolution, develops different escape mechanisms by presenting certain immune-suppressant ligands or by influencing its microenvironment by other means. Research on the interaction of immune cells with tumor cells led to the discovery of immunotherapy. Which escape mechanisms are present in the tumor, how are the immune cells responding to this, how do the immune cells interact with each other and what state are they in? These are essential questions for finding the right immunotherapeutic treatment for a given patient. What state are most of the macrophages in, M1 or M2, and is this influenced by the tumor? What is the concentration of the cytotoxic CD8-positive T-cells in the tumor and in its environment? How many regulatory T-cells are in the neighborhood of those cytotoxic ones with the potential to downregulate their activity? For every individual tumor those questions are answered differently, but it is clear that the decision for the right treatment depends on the set of those answers. Which set of answers is relevant for which treatment can be discovered by Tissue Phenomics.
Tissue Phenomics is the superior technology to collect cellular information
Tissue Phenomics is a technology that enables the generation and collection of the decisive information about cells and cell/cell interactions. Today, there is only one method available in the clinic that locally measures this most relevant information of cell/cell interactions inside a human body. This method is called histology which is the microscopic investigation of tissue slides including the staining of markers of certain proteins and genes therein. There is a long successful history of histology repeatedly making essential contributions to the understanding of biology and medical processes leading to the determination of the right treatment for patients. The most important successes happened in the field of oncology, and today, detailed and sophisticated analysis of tissue slides from cancer and other patients are the standard in the clinic. Tissue Phenomics is based on this histological knowledge and even goes beyond histology. The prerequisite is the digitization of tissue slides. Image analysis on digital slides enables a very rich quantification of the content captured therein. There might be one million cells or more in a tissue slide cut from a tissue block of a patient. Image analysis measures their position, shapes, types and states. This information can be collected from a single tissue slide with several markers that visualize different proteins, or it can be collected from several consecutive tissue slides from the same block with each of them being marked with just two or three markers. In the latter case, the consecutive slides can be combined with all the relevant information into one virtual slide. In the simplest case, the density of the different cells with their different states is collected for different regions. This is impossible to achieve with a manual approach, particularly if the quantification procedure is performed for many patients, like hundreds or – in a true big data approach – even millions. Even if a pathologist was able to create this massive amount of information, he or she would not be able to find patterns therein that predict clinical outcome. The amount of data points is immense and too big to be overseen for the discovery of patterns that correlate with clinical outcome.
Today, Tissue Phenomics is the only way of making this rich and valuable information available.
Tissue Phenomics creates valuable new medical knowledge
Valuable new medical knowledge can be created through Tissue Phenomics based on rich quantification on the cellular level that cannot be created otherwise. Most of the medically-relevant processes take place on the cellular level. The activation of molecular pathways within a cell as well as the interactions between cells represent local activities that are essential for living organisms. In histology, specific proteins can be marked and visualized, and those proteins can be used to either characterize the cell type or the state of the cell. This way, it can be concluded what types of cells in what states interact with one another. The dynamics of this interaction cannot be seen directly, but the proximities of cells and their densities in certain regions, the morphology of the cells and of their arrangements and their states in most cases tell a lot about their interactions. Cell states, morphology and cell/cell interactions can be quantified in a very rich form through image analysis.
In a subsequent data mining procedure, i.e. in sum through Tissue Phenomics, it can be investigated whether those behavior patterns represent fingerprints for predicting clinical outcome – in general or related to certain treatments. By now, it has already been shown many times that this can indeed be achieved with very high precision, although the field has just started.
Tissue Phenomics enables true personalized medicine
Tissue Phenomics adds a new dimension to profiling the diseases of patients, enabling a higher treatment precision and true personalization. Through Tissue Phenomics, it has been discovered that there is much more valuable information in medical images than had been anticipated. Particularly in oncology, there is more than one treatment option, and predicting what is the right therapy for a specific patient is essential. The meaning of the tremendously rich information in tissue can be unfolded through Tissue Phenomics and leading to a first step in significantly higher treatment precision.
Furthermore, future treatments will be more complicated than today when taking into consideration the complexity of the disease. Instead of one specific therapy, there will be sophisticated treatment workflows including sequences of treatments and combinations of treatments. This evolution has already started. Therefore, individual treatment workflows will be tailored with the help of Tissue Phenomics and other diagnostic tools very specifically to individual patients. This individualized interplay of profiling and treating patients will truly deserve the term personalized medicine.