[ARDD 2024 Recap] Dual-Purpose Target Discovery and Aging Research with Precious GPT Models

Summary

Frank presents the advancements in AI-driven drug discovery, focusing on dual-purpose target discovery for both aging and cancer treatments. He discusses the integration of multiomic data and AI models to identify potential therapeutic targets. The research highlights the discovery of the target TEIC for IPF (idiopathic pulmonary fibrosis), which progressed to clinical trials within 30 months. He also showcases the identification of the KD M1A target, validated through experiments, demonstrating how AI facilitates the discovery of targets that impact both cancer and aging simultaneously.

Takeaways

• 📈 Frank discusses the application of AI in drug discovery, specifically in target discovery, indication expansion, and drug repurposing.
• 🏥 The paper published in Translational Pharmacological Sciences highlights AI's role in drug discovery, with AI-identified targets advancing to clinical trials.
• 🧬 The AI and bioinformatics models utilize multiomic data to calculate disease-gene interactions in real-time, aiding in target discovery and indication expansion.
• 📚 A paper published in Nature about 'teic' as a target for IPF (Idiopathic Pulmonary Fibrosis) showcases AI's comprehensive impact on drug discovery.
• 🔍 The AI platform 'Pandemics' identified 'teic' as a promising target for IPF, emphasizing the importance of targeting age-associated diseases.
• 🧪 Generative AI, specifically 'Chemistry 42', was used to design novel molecules to target 'teic', demonstrating AI's role in molecular design.
• ⏱️ The development from target identification to clinical trials was achieved in a record 30 months, showcasing the efficiency of AI-assisted drug discovery.
• 🔬 The study integrates panomics to identify targets implicated in multiple aging diseases, considering aging as a primary driver of diseases.
• 🌐 Collaboration with the University of Oso focuses on identifying dual-purpose targets effective for both cancer and aging treatments.
• 🧪 KDM1A is highlighted as a potential dual-purpose target, being upregulated in multiple solid tumors and negatively associated with aging hallmarks.

Q & A

What is the main focus of the presentation?

The presentation focuses on the use of AI, specifically GPT models, in drug discovery, target discovery, indication expansion, and drug repurposing.

What success has been achieved since the use of AI in target discovery?

Since 2020, several AI-identified targets have been validated in the lab, and one of the identified targets has progressed to human clinical phase 2 trials.

What approach is used for target discovery in the research?

The approach uses multiomic data and text data, combined with AI and bioinformatics models, to calculate interaction scores between diseases, genes, and biological processes in real time. This helps identify target candidates for diseases or match the best indications for specific targets.

What specific program is highlighted as a success story in the use of AI for drug discovery?

The program highlighted is TEIC for idiopathic pulmonary fibrosis (IPF), which was identified as the best target using the AI platform Panomics in 2019. It successfully moved through preclinical trials and into clinical phase 1 and 2 within 30 months.

How does the research relate aging to drug discovery?

The research identifies therapeutic targets implicated in aging, as aging is seen as a driving factor for many diseases. The platform assesses how these targets are integrated into the hallmarks of aging to develop more effective treatments.

What is the significance of dual-purpose targets in this research?

Dual-purpose targets refer to targets that can be used to treat both cancer and aging simultaneously. The research aims to identify such targets by examining shared biological pathways in cancer and aging.

What is the significance of KDM1A in the study?

KDM1A is highlighted as a dual-purpose target, upregulated in 10 out of 11 solid tumors, and negatively associated with aging. Experiments show that knocking down KDM1A may increase lifespan, making it a promising target for both cancer and aging treatment.

What AI tool is used to design molecules targeting specific diseases?

The AI tool 'Chemistry 42' is used to design novel molecules that target specific diseases, such as TEIC for IPF, after the targets are identified.

How does the AI platform facilitate drug repurposing?

The AI platform facilitates drug repurposing by providing a list of the best indications for specific targets, allowing researchers to find new uses for existing drugs.

What collaboration is mentioned in the research regarding dual-purpose targets?

The research mentions a collaboration with Evandro from the University of Oslo, focusing on identifying targets integrated in both cancer and aging, particularly across 11 common solid tumors.

Outlines

00:00

🔬 Introduction to Dual Purpose Target Discovery and GPT Models

Frank opens by thanking the introduction and expressing excitement about discussing the dual-purpose target discovery and agent research using GPT models. He highlights a paper published in 'Trends in Pharmacological Science' that covers how AI assists in drug discovery, specifically target discovery, indication expansion, and drug repurposing. He notes the increasing validation of AI-identified targets, mentioning that one of their targets has progressed to human clinical trials (Phase 2). Frank briefly introduces their approach to using multiomic data and bioinformatics models to identify targets for diseases or discover the best indications for known targets.

05:00

💡 AI-Driven Target Discovery and Drug Repurposing Process

Frank dives deeper into their target discovery approach, using multiomic and tax data alongside 23 AI and bioinformatics models. These models calculate interactions between diseases, genes, and biological processes, allowing researchers to either start from a disease and find potential targets or start from a target and discover its best indications. He discusses a lead program, TEIC, identified as the best target for idiopathic pulmonary fibrosis (IPF) using their AI platform. The process, from identifying the target to designing new molecules and conducting preclinical and clinical experiments, has moved very quickly, with TEIC already in Phase 2 clinical trials in China and the US.

Keywords

💡Dual Purpose Target Discovery

Dual Purpose Target Discovery refers to identifying therapeutic targets that can treat multiple conditions, such as both cancer and aging, with a single drug. In the script, Frank discusses how AI models help in finding such targets that can be applied to various diseases, maximizing the efficiency of drug development.

💡AI in Drug Discovery

AI in Drug Discovery highlights the use of artificial intelligence to accelerate the identification of drug targets, improve drug repurposing, and expand indications. Frank emphasizes how their AI models assist in processes like target discovery and drug repurposing, making it possible to develop drugs more efficiently and accurately.

💡Target Discovery

Target Discovery involves identifying molecules (often proteins) that play a crucial role in the onset of a disease and can be targeted by drugs. Frank describes how their system analyzes multiomic and biological data to suggest the best targets for diseases, such as fibrosis and cancer.

💡Indication Expansion

Indication Expansion is the process of finding new therapeutic uses for an already existing drug. In the video, Frank explains how AI helps expand the indication of drugs by suggesting additional diseases that a drug might be effective against, based on its interaction with certain biological processes.

💡Drug Repurposing

Drug Repurposing refers to finding new therapeutic applications for existing drugs. In the script, Frank discusses how AI can identify potential new uses for drugs that have already been approved for different conditions, such as using an existing drug for fibrosis to treat aging-related diseases.

💡Multiomic Data

Multiomic Data integrates various types of biological data, such as genomic, transcriptomic, and proteomic information, to get a holistic view of biological processes. Frank explains how their AI models use multiomic data to compute the interactions between diseases, genes, and biological pathways, aiding in drug discovery.

💡Hallmarks of Aging

The Hallmarks of Aging are biological processes that are associated with aging, such as cellular senescence and intracellular communication. Frank discusses how their AI platform identifies drug targets that are implicated in several hallmarks of aging, offering a potential route to treat age-related diseases.

💡Clinical Phase II

Clinical Phase II is a stage in drug trials where the safety and efficacy of a drug are tested in a larger group of people after it has passed Phase I trials. In the video, Frank mentions that one of the AI-identified targets has successfully moved into human clinical trials, Phase II, a crucial step in validating the drug's potential.

💡KDM1A

KDM1A is a gene that has been implicated in both cancer and aging. In the script, Frank presents KDM1A as one of the dual-purpose targets that is upregulated in tumors but downregulated in aging, and he explains how their AI model suggests that modulating KDM1A could help extend lifespan while also treating cancer.

💡Generative AI Chemistry

Generative AI Chemistry refers to the use of AI to design novel chemical compounds that could serve as new drugs. Frank discusses how their AI platform, Chemistry42, was able to design new molecules to target specific diseases, speeding up the drug discovery process by generating potential drug candidates automatically.

Highlights

Introduction of Dual Purpose Target Discovery and agent research using GPT models.

AI aids in drug discovery, including target discovery, indication expansion, and drug repurposing.

Paper chosen as Editor's Choice and published in Trans Pharmacological Science.

AI-identified targets validated in the lab, with one moving to human clinical phase two trials.

Use of multiomic and tax data, along with 23 AI and bioinformatics models, to calculate disease and gene interactions in real-time.

AI platform helps generate target candidates for diseases or identify best indications for specific targets.

Published paper on TEIC for IPF in Nature Tech Journal, showcasing AI's impact on drug discovery.

Identified TEIC as a target for IPF using the AI platform, with confirmation of its role in several hallmarks of aging.

Generative AI used to design novel molecules for the identified target, resulting in rapid progress to clinical trials.

TEIC progressed from target identification to clinical phase one within 30 months, a record pace.

AI identifies therapeutic targets linked to aging, with focus on multiple aging diseases.

Collaboration with University of Oslo researchers on cancer and aging treatment using AI.

Discovery of dual-purpose targets that address both cancer and aging simultaneously.

KDM1A identified as a promising dual-purpose target for cancer treatment and lifespan extension.

Successful RNA experiment in C. elegans showing knockdown of KDM1A increases lifespan, showcasing AI's role in identifying dual-purpose targets.

Transcripts

thank you very much for the introduction

of Alex and hello everyone uh I'm Frank

I'm really excited today uh to talk

about our work on the Dual Purpose

Target Discovery as well as the agent

research with our precious uh GPT

models um yes this is the paper that we

published last year in Trans in

pharmacological science that we talk

about how AI can help um in the drug

Discovery um including for example the

uh uh Target Discovery indication

expansion and Drug repurposing okay and

the paper was also chosen by the as the

addit choice last year so since 2020 we

can see that more and more AI identified

targets has been being validated with v

lab and one of the targets has been even

moved to the human clinical phase two

that is our teic targets so it's a very

uh exciting okay okay and then this is

our inal approach how we do the target

Discovery so basically we use multiomic

data and tax data um we use our 23 Ai

and bio informatics models to calculate

the score for the interactions of the

disease and genes and the biological

process in real time so that let's say

we can start from the disease of

interest and then the system will help

us to provide a list of of the target

candidates um for the best Target for

the disease or we can start from the

target of interest and then this to Pro

the system will provide a list of the

best indication for that particular

targets so that is the way how we do the

uh Target Discovery indication expansion

and drw

repurposing so early this year we

published a paper uh that is our lead

program the teique um for ipf on the

nature about Tech Journal so uh as far

as I know it is the most successful and

the most comprehensive story example to

show how AI can really help uh drug

Discovery so you can see that since

2019 we use our AI platform pandemics to

identify teic as the best Target for ipf

and since ipf you know the fibrosis is

kind of a age associate diseases so

that's why we also do the hmark of Aging

assessment it turns out teic has been

inegrated in several Hallmarks of Aging

for example the Ed uh intracellular

Communications and cellular

senz so once we confirmed the targets

and then we use our generative AI uh the

chemistry 42 to design a truly noal um

molecules to tackle the this Target and

then we went through a series of the uh

invitro inval uh preclinical experiment

and then we moved to the clinical within

30 months so this is an other industry

record U that it move very fast and now

we already finished the phase one and we

have moved to phase two in both China

and

us so since we identified teique in 2019

as the best Target for ipf and then we

truly believe uh finding the therapeutic

Targets implicated in aging is very

helpful so that's why we try to use our

panomics to very uh course effectively

to find the targets integrated in

multiple aging disease because we

believe that aging is the driving force

for most of the diseases and then we can

also see tenix is implicated in several

hormones of Aging here and uh after we

identify the targets for multi um age so

disase we further classify them uh into

different horor of

aging and then uh last year we building

on this momentum um we focus on whether

we can find the DU Pro Target uh for the

treatment of cancer and aging so that's

a very interesting so um we collaborate

with evandro from University of Oso to

focus on this study so basically we try

to identify the target um that is

commonly uh integrated in 11 solar

tumors um those are the common cancer

targets and then we also used the uh GTS

half RN SE data to find the targets

associated with aging and then

we um classify those Target into several

groups so you can see that the green

targets are those uh with evidence in

extending lifespan so here our

definition of the D Target is that uh

with the same direction of interaction

that we can achieve the goal for the

treatment of cancer at aging at the same

time so probably you know the best

example is the Amor uh the rber meing

example so however you can also see that

there's a over 100 U cancer targets that

do not have uh the known effect on

lifespan so next uh we are going to

propose uh some uh D propose targets at

that we propose tr3 of them and the most

interesting one is the KD M1A so from

this figure from panel a you can see

that KD M11 a is uppr regulated in 10

out of 11 solid tumors at the same time

uh kdm1a is negatively associated with h

i mean these Expressions uh so you can

see that kdm1 is is down regulated so uh

I'm not surprised because you know Mt

also have the similar situation so

that's why we want to see whether the

knockdown of kdm1a can really help to

incre increase the LIF span so that's um

we did the uh RNA uh Northern experiment

in the sea elegant and the result is

very impressive so it is a very good

example to show that how AI can

facilitate the identification of the DU

purose targets so uh yeah this is uh uh

the the story I want to share today and

now I will pass to my colleag

uh he will introduce a very interesting

uh gbd models thank you