Quality Alchemist (品質煉金術師): CityU Distinguished Lecture on AI Enabled Personalized Theranotics

2019年1月28日星期一

CityU Distinguished Lecture on AI Enabled Personalized Theranotics

To enhance academic ambience and stimulate intellectual vitality, CityU invites world renowned scholars, high calibre figures of the community and social elites to deliver Distinguished Lectures on a wide range of interesting topics to the university community and members of the public. I attended the lecture named “AI Enabled Personalized Theranotics” on 28^th Jan 2019. I also had a chance to take a photo with Prof. Chih-Ming Ho for memory.

In the beginning, Prof. Jian Lu (VP (Research and Technology), CityU) introduced Prof. Chih-Ming Ho background and achievement to us. Professor Chih-Ming Ho received Ph.D. from The Johns Hopkins University. He held the Ben Rich-Lockheed Martin Chair Professor until retired in 2016 and currently is a UCLA Distinguished Research Professor. He served as UCLA Associate Vice Chancellor for Research from 2001 to 2005.

Prof. Chih-Ming Ho (Distinguished Research Professor, University of California, Los Angeles; Member, US National Academy of Engineering; Academician, Academia Sinica) started his talk to share AI Enabled Personalized Theranostics. Firstly, he acknowledged his team in tens of collaborators’ laboratories included 50 PhD and 150 visiting scholars. Then he explained that “Medicine is a Science of Uncertainty and an Art of Probability” and briefed the Father of Modern Medicine Sir William Osler (1849-1919).

Then Prof. Ho mentioned Theranostics (Therapeutics plus Diagnostics) could be considered as engineering term from input to output. One of challenges is that molecular mechanisms could not provide guidance of dose levels that related to toxicity management. His study the question that what is the Transfer Function from input “Drug & Dose” to output “Efficacy / Toxicity”?

And then he introduced that almost all diseases treated by combinatorial drug such as AIDS treated by three drug combination after 1995 and the death per year reduced from around 50,000 to 20,000 and then stable to 2013. However, he also found that almost all combinatorial drugs were not optimized.

Prof. Ho stated the challenge for Personalized Clinical Applications from Drug and Dose to “Big Data” to Phenotype. Small Data means clinician has only two actionable parameters (i.e. Drug and Dose). Big Data means Physiological functions and cellular omics, as well as, optimized drug-dose combination test.

In order to solve this challenges, Prof. Ho employed Artificial Intelligence (AI) that bypassed the mechanism-based approaches and was a mechanism-free method. AI could be bypassing the big data and bridging input and outputs. Prof. Ho shared discovery of the AI-Phenotypic Response Surface (AI-PRS) Equation to model the drug-dose combination.

After that Prof. Ho selected some examples such as TB Drug Combos from a large pool (14 drugs), Chemotherapy for cancer, PRS Based Colon Cancer Preclinical Test, etc. The last example was AI-PRS guided personalized dosing in a metastatic prostate cancer patient and it was found that to reduce one of dose from the range 36 - 48mg to 24mg. (Remark: Doctors usually did not agree with it!)

Finally, Prof. Ho discussed the challenges of Personalized Diagnosis included “Type of Disease”, “Measuring the therapeutic efficacy” and “Relapse of a disease”.

Lastly, he concluded that think differently and life could be very simple. Do not make it too complex!

Q&A Session

Prof. Michael Yang responded to Prof. Ho’s conclusion. He said they worked in molecular mechanism was made it as complex as possible that was opposite to AI-PRS Platform.