21 Dec PMWC Interview with Janusz Dutkowski, Ph.D., Co-founder & CEO, Data4Cure – Speaker at PMWC 2018 Silicon Valley
[et_pb_section admin_label=”section” transparent_background=”off” background_color=”#ffffff” allow_player_pause=”off” inner_shadow=”off” parallax=”off” parallax_method=”off” custom_padding=”0px|||” padding_mobile=”on” make_fullwidth=”off” use_custom_width=”off” width_unit=”on” make_equal=”off” use_custom_gutter=”off” custom_padding_tablet=”-30px|||” custom_padding_last_edited=”on|desktop”][et_pb_row admin_label=”row” make_fullwidth=”off” use_custom_width=”off” width_unit=”on” use_custom_gutter=”off” custom_padding=”0px|||” padding_mobile=”on” custom_margin=”0px|||” allow_player_pause=”off” parallax=”off” parallax_method=”off” make_equal=”off” parallax_1=”off” parallax_method_1=”off” parallax_2=”off” parallax_method_2=”off” column_padding_mobile=”on” custom_padding_tablet=”0px||-50px|” custom_padding_last_edited=”on|desktop” custom_margin_tablet=”0px||-50px|” custom_margin_last_edited=”on|phone”][et_pb_column type=”4_4″][et_pb_accordion admin_label=”Accordion” use_border_color=”off” border_color=”#ffffff” border_style=”solid”]
[et_pb_accordion_item title=”Q: Tell us about Data4Cure.”]
A: Data4Cure develops and provides its partners with an integrative platform that combines systems biology, machine learning and AI technologies to turn vast amounts of omics, phenotypic and clinical data (both public and proprietary) into a data-driven biomedical knowledge graph. The platform uses this information to build comprehensive maps of diseases to identify new drug targets and biomarkers, find new indications for existing drugs, and match drugs to patient subtypes. We are proud to work with some of the largest and most innovative pharmaceutical companies and leading research institutions who use our platform to gain new insights from their data and build their internal knowledge networks.
[/et_pb_accordion_item][et_pb_accordion_item title=”Q: What led you to found Data4Cure?”]
A: As a computer scientist and computational biologist, I am always interested in how we can combine and model different types of data together to better understand complex systems and make better decisions. Around 2013, I was working at UC San Diego, developing and publishing on new network and systems approaches which worked well for simple organisms such as yeast but weren’t yet as effective for human disease. I thought we can do better by bringing in a lot more diverse types of data and merging systems biology with some of the new machine learning and semantic learning technologies that were becoming available. That would require a new set of tools and a computational platform that would start to tie all of the pieces together. I thought the work could be more effective done through a company. So instead of looking to start my own academic lab, I left to start Data4Cure helped by our scientific co-founder Dr. Trey Ideker, key scientific advisors Drs. Lee Hood and Napoleone Ferrara, as well as early investors. We were very fortunate to be joined by extraordinary computational biologists and engineers and began working with pharmaceutical companies to apply some of the integrative systems biology machinery toward patient stratification and drug development. It gave us a good view into the problems people are facing both in terms of specific analytical challenges, as well as the bigger problem of growing knowledge from more and more data.
What we were able to develop over the last three to four years is an integrated system that not only applies many advanced analytical methods to the data but provides a technology platform to iteratively update knowledge based on new data and literature. The resulting information is stored in a dynamic data-driven knowledge graph that is unique to each organization. This technology became the basis for the Biomedical Intelligence Cloud — our SaaS offering for pharmaceutical R&D which we launched last year. We are continuously improving it by curating and feeding in new large datasets and building new applications to address a wider range of analytical problems.
[/et_pb_accordion_item][et_pb_accordion_item title=”Q: What are some of the key unique ideas behind the Biomedical Intelligence Cloud? What contextual knowledge do you bring in?”]
A: The key idea is convergence of evidence and information coming from multiple sources like multidimensional omics data, phenotypic data, molecular networks and pathways, literature, clinical trials, and other structured and non-structured data that may exist both within the organization and outside in the public domain. Typically information coming from each one of these sources is noisy, incomplete, and difficult to interpret. By putting the data together in context we are able to extract the most robust pieces of information and make better sense of it.
The Biomedical Intelligence Cloud uses a proprietary knowledge graph technology called CURIE to continuously aggregate various pieces of evidence, learn how they fit together and apply them toward interesting problems in the pharmaceutical and clinical R&D. We think of it also as convergence of systems biology and ML/AI because the technology leverages recent methods from machine learning and AI but is also heavily informed by the developments in systems biology and incorporates a lot of prior knowledge from the biomedical domain.
[/et_pb_accordion_item][et_pb_accordion_item title=”Q: What types of problems in pharmaceutical R&D do you help address and how does systems biology and ML help?”]
A: The platform has applications throughout the drug development pipeline from early target discovery and validation, to mechanism of action and efficacy studies, to patient selection and developing companion diagnostics. In each of these areas there is an opportunity for integrative systems-based modeling but there are also unique challenges. The platform helps address these individual areas using a set of dedicated applications.
For instance, to support target discovery and validation, users might start by building integrative molecular maps of diseases which integrate omics, phenotypic and literature information with molecular networks and pathways to identify the key system components with potential for therapeutic intervention. Precision medicine is all about precisely mapping disease subtypes and predicting how they might respond to therapy either as single agent or as combinations. This is where some of the other applications become useful. We have an application that performs a deep molecular subtyping of disease based on multidimensional omics data and specific disease pathways perturbed in each patient. Other applications couple machine learning algorithms with knowledge of molecular networks and pathways to predict sensitivity or resistance to drugs in specific patient subsets. We currently have over 10 applications available on the platform and are expanding with new ones in the next months.
[/et_pb_accordion_item][et_pb_accordion_item title=”Q: You will be speaking in the Immunotherapy Track as well as the AI showcase at PMWC 2018. What are the applications of the platform toward immune oncology? Can you give some use case examples of cancer biomarkers that the Biomedical Intelligence Cloud helped discover?”]
A: Immuno-oncology research and clinical trials is one the areas in which our platform has seen the most growth in 2017. We are particularly interested in understanding and predicting which patients across different histologies will respond to specific types of immunotherapy (e.g. PD1 or PD-L1 inhibitors) and which other patients may respond to combination therapies in which IO drugs might be combined with other therapies.
The power of the platform is that we can effectively combine multiple types of data, such as are often generated by the new IO clinical trials and apply a range of systems and ML approaches to make sense of these data. What is emerging is a view of the tumor-immune system interactions in which multiple factors such as cancer subtypes, immune infiltration, neoantigen load, checkpoint expression and tumor clonal evolution each may together determine response.
Our system can extract multidimensional information about disease biology coming from the analysis of large research cohorts and use it to boost signal and increase the ability to make accurate predictions for new clinical trial cohorts. In my talk I plan to discuss several case studies using recent trials in lung adenocarcinoma and melanoma.
We are very excited to contribute to the PMWC 2018 program. We’ve been part of the conference since the very beginning of Data4Cure. Every year the quality of the program and level of participants makes it a phenomenal event for anyone working in the field. The 2018 edition will be particularly interesting for anyone interested in immuno-oncology, and applications of machine learning and AI in medicine.
[/et_pb_accordion][/et_pb_column][/et_pb_row][et_pb_row admin_label=”Row” make_fullwidth=”on” use_custom_width=”off” width_unit=”on” use_custom_gutter=”off” custom_padding=”-50px||-50px|” padding_mobile=”on” custom_margin=”-50px||-50px|” allow_player_pause=”off” parallax=”off” parallax_method=”off” make_equal=”off” parallax_1=”off” parallax_method_1=”off” column_padding_mobile=”on” custom_padding_tablet=”-100px||0px|” custom_padding_last_edited=”on|desktop” custom_margin_tablet=”-100px||0px|” custom_margin_last_edited=”on|desktop” disabled=”off” disabled_on=”off|off|off”][et_pb_column type=”4_4″][et_pb_divider admin_label=”Divider – For Spacing” color=”#e2e2e2″ show_divider=”on” divider_style=”solid” divider_position=”top” hide_on_mobile=”off” custom_css_main_element=”max-width: 1200px;||margin: auto”]
[/et_pb_divider][et_pb_text admin_label=”Text – Inserted Blog” background_layout=”light” text_orientation=”left” use_border_color=”off” border_color=”#ffffff” border_style=”solid” custom_margin=”-50px|||” custom_padding=”-50px|||” custom_margin_tablet=”-125px||-125px|” custom_margin_last_edited=”on|tablet” custom_padding_tablet=”-125px||-125px|” custom_padding_last_edited=”on|tablet” disabled=”off” disabled_on=”off|off|off”]
[insert page=’blog’ display=’content’]