Lecture 1 notes Flashcards
Two things when developing a drug or giving a drug to a patient are important efficacy and safety. You want the drug to be ________and _______
effective and safe
Without pharmacokinetic analysis and data those safety and efficacy are just two words. In others having a concentration from blood at any time point without knowing the models without knowing the mathematics and principle of pharmacokinetic is useless and as we go your going to understand why…we cant just take random blood sample from patient and then try to adjust the dose because you need the knowledge of the model and how the drug behaves overtime
Estimation of PK parameters are pivotal because if you don’t estimate the parameters of PK then you cannot plug them into the model and essentially calculate the dose and frequency
Remember DOSE AND FREQUENCY
Many of PK knowledge is generating every year so therefore its essential to develop the ability to go over papers, analyze them, and extract the relevant data
Generative AI is fundamentally changing drug development and precision medicine
AI can look at sparse data and with 100% precision tells you specifically what is the best dose for this particular patient based on the data it receives…so you train the model and the model actually tells
Future consequences of Model-Informed Drug Development is that’s its going to reduce cost and time for development. ((Currently it takes 15 years for drug development and with MIDD it may only take half that time…its going to save lots of money for companies…why?…because they are going to use the digital twin for simulation so instead of using real patient or real subject for the clinical phases they are going to use virtual patients and those virtual patients are going to surrogate for real projects…so instead of recruiting 5000 people for a vaccine you can a million digital twin as precise as real person and you can run simulation in one two day and then you get very precise prediction of the clinical performance of the drug and that is the whole goal of having MIDD digital twin
PK/PD model depends on differential equations or some rate constant and at the end you will come up with a model that helps you support your decision
(90% of drugs enter clinical trials and they fail but with MIDD it could probably only 50% of drugs fail
FDA and EMA are holding a lot of sessions to support the MIDD
They are ways they are using patient sample ex vivio in 3D microfluidic chips generating the data and then put them in this model and models can predict very precisely what is going to happen in clinical setting
When talking about digital twins you are going to use the principle of MIDD
(You can utilize the power of statistical approaches, machine learning and generative AI to incorporate genomics into your model as a covariate
(Every day there is few publication on new model or improving QSP models, quantitative pharmacology models that better describe the mechanistic or mechanism of action of drugs, cellular level, intercellular level, peripheral central compartment,etc
Each time there is an interaction between the model and the user the generative AI is going to learn and going to reduce the mistakes and the next time when it approaches to do the computation its going to refine the parameters and give you more precise answer.
There are ways you can make MIDD hallucination free to give you 100% precise data
SIM man at western new England university is a simulation patient but it will get more precise its constantly learning and constant monitoring to that.
Quantitative Systems Pharmacology is key for improving preclinical translational science to have a more precise in human prediction…we are moving from traditional animal studies and traditional trial and error toward more precise model informed generative AI insightful drug development
Do sensitivity test to see what parameter is more relevant to your model then optimization like linear correlation to optimize drug then your going to do another simulation now you have a validated calibrated model which can then be used for predicting other scenarios
Lets say you want to predict another dose…you tried 40 microgram now you can 100 microgram…the importance of this many of the time that clinical trial fails is because choosing wrong dose and frequency
