Title: Scientist’s computational lawyer
Abstract:
I will review the tangible improvements my colleagues and I have achieved in novel biomarkers across various clinical specialties, and which have led to more accessible, affordable, and accurate diagnostics. Partially, these advances are due to non-trivial computational analyses and ingenious defense of the scientific hypotheses. I will outline the credo of a scientist’s "Computational Lawyer." A skilled one can rescue a flawed experimental design and successfully defend a scientific hypothesis, even when dealing with sub-optimally collected data. Should an analyst use AI? Yes, but only in a manner commensurate with their competence. It does not imply that a data scientist or bioinformatician can outsource her thinking or compensate for any deficiencies in competence with a magic black box of AI, even though she may believe it is reliable. The AI era demands more knowledge, not less. It fundamentally requires understanding the algorithms behind the employed R or Python libraries, as well as relatively new phenomena, for example information leakage.
