emerging technologies: implementing ai, ml, and dl to drive … · 2020. 12. 9. · eer teles pleet...

9
Emerging Technologies: Implementing AI, ML, and DL to Drive Drug Discovery New technologies are playing an increasingly important role in bio-pharma research.

Upload: others

Post on 30-Dec-2020

8 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

1EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

Emerging Technologies: Implementing AI, ML, and DL to Drive Drug DiscoveryNew technologies are playing an increasingly important role in bio-pharma research.

Page 2: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

2EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

Drug discovery is an incredibly costly and time-consuming process. There are millions of potential compounds out there, and researchers now have unlimited possibilities for de novo drug design.

Qualifying enormous numbers of possible drug candidates requires

computing infrastructure that only the most sophisticated bio-pharma

research labs possess. Among those that do, there are even fewer that

have integrated efficient, automated strategies for using these resources

in an optimal way.

Computational-aided design is nothing new for bio-pharma researchers,

but today’s cutting-edge technologies offer transformative benefits far

beyond what was possible mere years ago. The key to leveraging these

benefits effectively lies in successful integration, informed by expert

guidance that treats the laboratory as a unified whole.

Page 3: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

3EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

Three Emerging Technologies and Their Drug Discovery Research Applications

Research laboratories around the world are focusing on three key

technologies to overcome production bottlenecks and improve research

outcomes while reducing costs. These technologies represent some

of the most important advances that researchers can leverage when

performing drug development tasks.

• Artificial Intelligence (AI)

Artificial intelligence is a broad discipline that encompasses

automated decision-making and pattern-recognition. It includes

several subset disciplines, including machine learning, computer

vision, and natural language processing.

• Machine Learning (ML)

Machine learning is a subfield of AI that focuses on algorithms that

can learn without human intervention. When these self-learning

systems identify patterns in data sets, they qualify their results to

become better at identifying those types of patterns. There are

three main types of ML algorithms: supervised, unsupervised, and

reinforcement learning.

• Deep Learning (DL)

Deep learning is a subset of machine learning that focuses on

very large data sets. These systems learn from example and filter

out statistical noise to glean insight in ways humans can’t easily

replicate. Convolutional Neural Networks, Recurrent Neural

Networks, and Recursive Neural Networks are three examples of

DL algorithm architectures.

Page 4: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

4EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

Each of these disciplines presents a unique set of challenges to bio-IT

research professionals and drug development scientists. Researchers

who are invested in data infrastructure will need to overcome these

challenges with expert guidance.

Artificial Intelligence: Breaking Down Barriers to Clinical Research

Artificial Intelligence presents some of the most exciting opportunities for

bio-IT investment and drug research acceleration. Researchers around

the world are using AI-powered scientific computing engines to support

clinical decision-making processes and optimize drug research.

Since AI is such a wide discipline, bio-pharmaceutical professionals are

using it in many different ways. The sheer number of AI-powered options

and use cases is one of the greatest challenges that bio-IT teams have

to face.

RESEARCHERS AROUND

THE WORLD ARE USING

AI-POWERED SCIENTIFIC

COMPUTING ENGINES

TO SUPPORT CLINICAL

DECISION-MAKING

PROCESSES AND OPTIMIZE

DRUG RESEARCH.

Page 5: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

5EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

POWERED BY ELECTRONIC

HEALTH RECORDS AND

WEARABLE MEDICAL

DEVICE DATA, AI ALSO

HAS THE POWER TO

AUTOMATE THE PROCESS

OF CONNECTING

ELIGIBLE CLINICAL TRIAL

VOLUNTEERS WITH THE

SCIENTISTS RUNNING

THEM.

Bio-IT executives know that AI can simplify complex systems and cut

down research costs. The problem is that they only have finite time and

resources, and cannot afford to implement AI-powered processes across

the board—nor should they.

Instead, bio-pharmaceutical executives need to identify the few areas

where AI-powered transformation can have the greatest impact on

research outcomes. Laboratories that focus their integration efforts on

their highest-value processes will earn the greatest long-term benefit

while incurring the least infrastructural risk.

Many bio-pharmaceutical companies have already identified these

AI-ready processes. One of the areas where AI consistently delivers

improvement is in clinical trial recruitment. Until now, if researchers

wanted a standardized, indexable database of eligible trial patients, they

had to make one themselves —and learn a database query language like

SQL to use it.

AI has the power to transform drug manufacturing processes on multiple

levels. The ability to glean insights from AI-powered simulations and

modeling can help researchers focus on the most promising compounds

and pharmacodynamic methods.

Powered by electronic health records and wearable medical device

data, AI also has the power to automate the process of connecting

eligible clinical trial volunteers with the scientists running them. This will

save billions of dollars, improve trial success rates, and even address

problematic selection biases, like the fact that 79% of genomic data

comes from a demographic that only represents 16% of the

world’s population.

Page 6: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

6EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

Machine Learning Offers Fast, Accurate Predictions

Machine learning holds great promise in the world of automated

decision-making, personalization of drug therapies, and clinical data

governance. The ability for machine learning algorithms to improve the

accuracy of their insights over time makes them ideal for establishing

analytical roadmaps for the drug discovery process.

Unsupervised ML techniques can predict the therapeutic efficacy of

known and unknown pharmaceuticals. These systems can also play

an important role in predicting the outcome of drug repurposing trials,

and help researchers interpret the molecular mechanisms of

different compounds.

One of the ways that ML can achieve this is by grouping compounds

based on gene expression similarities and clustering the compounds that

have mechanisms of action and biological pathways in common. This is

the essential promise of MANTRA 2.0, developed by the di Bernardo Lab

of the TeleThon Institute of Genetics and Medicine.

Page 7: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

7EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

ML-powered technologies can also reduce the cost of toxicity prediction.

Advanced self-learning algorithms can identify similarities among

compounds and predict their toxicity based on input features. eToxPred

uses machine learning to estimate the synthesis feasibility and toxicity of

small organic molecules with promisingly high accuracy.

As with artificial intelligence, research executives and Bio-IT teams need

to work together to identify the best areas to incorporate machine

learning into the drug discovery framework. Particular attention must be

paid to the self-learning nature of machine learning, which relies heavily

on the availability of accurate, well-structured data to derive insight.

This means that some processes are better-suited to machine learning

simply by virtue of having more comprehensive data sets available. High-

quality drug discovery data suitable for machine learning use is relatively

limited in quantity, and must pass stringent validation tests before

entering the drug discovery workflow.

Deep Learning Streamlines Drug Discovery Insights

Deep learning relies on artificial neural networks to simulate the way the

human brain processes information. This typically requires the large-scale

deployment of interconnected computing elements that function in a

way analogous to biological neurons. By mimicking the transmission of

electrical impulses in the brain, neural networks can identify patterns and

solve problems that other technologies cannot.

Since deep learning algorithms typically rely on huge data sets and vast

computational spaces, they are ideally positioned to address some of

the most fundamental problems that bio-pharmaceutical researchers

often face.

Page 8: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

8EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

Deep learning engines can recognize hit to lead compounds over a far

wider query space than a human researcher. They can speed up drug

target validation and even help with drug structure design optimization.

Some of the applications for deep learning in drug discovery

include:

• Predicting the 3D structure of target proteins

• Predicting drug-protein interactions

• Determining biospecific drug molecules

• Designing multi-target drug molecules

• Predicting bioactivity and physicochemical properties for drug screening

Deep learning predictions can help researchers identify the best

direction to focus further research. With DL-powered solutions onhand,

researchers will spend far less time traveling down dead-end bio-

pharmaceutical paths.

WITH DL-POWERED

SOLUTIONS ONHAND,

RESEARCHERS WILL

SPEND FAR LESS TIME

TRAVELING DOWN

DEAD-END BIO-

PHARMACEUTICAL PATHS.

Page 9: Emerging Technologies: Implementing AI, ML, and DL to Drive … · 2020. 12. 9. · EER TELES PLEET A, L, AD DL T DRE DR DSER 1 Emerging Technologies: Implementing AI, ML, and DL

9EMERGING TECHNOLOGIES: IMPLEMENTING AI, ML, AND DL TO DRIVE DRUG DISCOVERY

In order to capitalize fully on the predictive power of deep learning,

research scientists will need to identify accurate, high-impact data sets

to target with DL-powered tools. In many cases, data validation will be

necessary to improve overall structure and consistency before running

DL tools.

It’s also important for researchers to pay attention to the economics

of deep learning. Neural network training and operation is resource-

intensive, and comes at a significant cost. DL-powered analyses of very

large systems—on the order of 1.5 billion parameters—can cost between

$80,000 and $1.6 million to train.

This means research organizations need to carefully select the highest-

impact field for deep learning prediction, and pool resources to gain

optimal access to cloud-enabled deep learning infrastructure.

Integrate Next-Generation Technologies in Your Research Lab with Expert Guidance

RCH Solutions is ready to help you identify the research areas best-served

by emerging AI, ML and DL technologies. Our expert consultants can help

you optimize infrastructural investment and leverage new tools in the

most efficient way. Speak with one of our team members to begin the

process of optimizing drug discovery for your research organization.

RESEARCH

ORGANIZATIONS

NEED TO CAREFULLY

SELECT THE HIGHEST-

IMPACT FIELD FOR DEEP

LEARNING PREDICTION,

AND POOL RESOURCES

TO GAIN OPTIMAL

ACCESS TO CLOUD-

ENABLED DEEP LEARNING

INFRASTRUCTURE.

ABOUT RCH SOLUTIONS RCH Solutions (RCH) is a global provider of computational science expertise, helping Life Sciences and Healthcare companies of all sizes clear the path to discovery. For nearly 30 years, RCH has provided focused experience and unmatched specialization designing and deploying cross-functional IT strategies, supporting R&D infrastructure, and offering workflow best practices that solve enterprise and scientific computing challenges.

rchsolutions.com | [email protected]