1 of 18

AI in Drug Discovery: where are we really?�

Andrea Beccari, Senior Director EXSCALATE at Dompé Farmaceutici SpA

Soundarya Palanisamy, Sr Industry Data Scientist, SAS

SAS Corporate

2 of 18

Accelerating the discovery

of novel medicine with

EXSCALATE

“Model to Cure”

Andrea R. Beccari, PhD

Senior Director, EXSCALATE

Dompé Farmaceutici SpA

3 of 18

Has a holistic approach that combines computer simulations with machine learning to expand the existing experimental knowledge with high-quality synthetic data

EXSCALATE

4 of 18

EXSCALATE: Artificial Intelligence (AI) and Molecular Simulation (MS)

Combine complementary approaches to maximize the quality of predictions

Ligand Based Models

Molecular Docking

Safety Prediction

Binding Sites

Protein – Protein Interactions

Metabolism

4

5 of 18

High Performance Computing in Pharma Research

Data Explosion in Drug Discovery

5

Biology

Simulations

Chemistry

Drug Profiling

Scientific Knowledge

0.2 Million

200 Million

10 Million

>1 Trillion

0.1 Billion

36 Billion

~1K

28 Million

30 Million

Health Care Data

Billions

Corporate Data

In the last 5 years, we have seen an order-of-magnitude increase in the size of datasets.

This is forcing us to redesign our approach to drug design and data modeling.

5

6 of 18

PropheX: Comprehensive collection of model for drug’s safety assessment

Model Explosion

>600 pharmacologically relevant targets, 3 mechanism of actions and >10 AI algorithms: thousands of models to be stored, versioned, ensembled in consensus, deployed, updated and enhanced.

325 kinases

190 GPCRs

>600

Modelled targets

> 10

Liabilities

>10

ML techniques

Cardiotoxicity

Neurotoxicity

Hepatotoxicity

Metabolism

Endocrine tox.

Gastro. toxicity

Nephrotoxicity

Immunotoxicity

>2000

Validated Models

6

7 of 18

PropheX: Comprehensive collection of model for drug’s safety assessment

Model Life Cycle

ProfhEX platform fully automates the Model Life Cycle process, from data management to forecasting.

ProfhEX is accessible via secure APIs & webservices supporting the decision-making process

Model Life Cycle

External applications & Reporting

7

8 of 18

Speed to Results

How SAS Viya is a Productivity Multiplier for Life Sciences R&D

9 of 18

Drug Development Life Cycle

Drug Discovery QSAR

Precision Medicine for Pandemic Preparedness

Drug Repurposing with Network Biology

Optimizing Cell Manufacturing Quality

LEAD

GENERATION

LEAD

OPTIMIZATION

PRE-CLINICAL

STUDIES

CLINICAL

TRIALS

POST-MARKET

SURVEILLANCE

HEALTHCARE

DELIVERY

10 of 18

Drug Discovery and QSAR

BINDING AFFINITY PREDICTION

Absorption

Distribution

Metabolism

Excretion

Toxicity

QSAR = Quantitative Structure Activity Relationship

ADMET

USE CASES

1

2

Discovery

10x savings

10x speed-up

11 of 18

BENEFITS

Drug Repurposing: Path to Value

Out-of-the-box preliminary results

Refine with additional data

Augment with additional methods

Drug rescue

Disease endotyping

Personalized medicine

Drug combinations

Discover therapeutics for

rare or new diseases

Discover new targets for your existing

drug portfolio

Shorten Time

to Market

Lower Drug

Development Costs

Map potential

Off-label uses

DRUG REPURPOSING

Repurposing

12 of 18

Optimizing Cell Manufacturing Quality�

Data Integration
Multimodal ML/AI
Optimization

REGENERATIVE MEDICINE

Decision / Control

Optimizing

Modular Cell Production

Cell Imaging

Single-Cell Sequencing

13 of 18

Transforming treatment of viral-induced diseases

Precision Medicine

PRECISION MEDICINE

Precision Medicine

14 of 18

SAS Viya as a Productivity Multiplier

Optimizing Data Management, Maximizing Transparency and Accelerating R&D

15 of 18

Offline Dataset

External Database/File
Server Connection

S3, GCP, Azure Blob, Snowflake, Oracle, PostgreSQL, SingleStore, Redshift

Healthcare specific adaptors:

FHIR, OMOP, SAS health CDM

Synthetic Dataset - Courtesy of Scott McClain

CSV

JSON

sas7bdat

XPT

Parquet

Transmitted via:

REST API call�proc http
Electronic Data Capture system (EDC)
--> Medidata Rave, Veeva

SAS® Data Studio – Prepare Data

SAS® Data Explorer – Manage Data

16 of 18

SAS® Visual Analytics – Explore and Visualize

Model Studio – Build Models

17 of 18

PRODUCTIVITY

REPRODUCIBILITY

TRANSPARENCY

Combine code-based and low-code nodes together in one pipeline

Studio Flow

Notes: Each node executes one at a time, and you can manage the order of execution using swim lanes.

Capabilities:

Use SAS & Python Code, Directly Include Tables, Import/Export Data, Implement Custom Steps, Transform Data

BIG THREE:

Productivity

Because SAS Studio Flow relies on input from previous nodes to run, users can easily expand, restrict, or fine-tune other parts of the process without other nodes being negatively affected, as long as the component nodes of the workflow are reconnected. The clarity and ease of use allows users to be more productive in their data processing efforts

Reproducibility

When using macro variables, the ability to replicate the flow for different uses is high because the user only has change one item at the beginning of the flow and set it to run

Transparency

UI-based nodes make the flow interactive and clear. Code-based nodes can be thoroughly commented. The user can also create notes along the flow. Should any issues errors occur, SAS Studio Flow allows the user to immediately identify the affected node, making debugging a vastly more achievable task than in a longer SAS program. Users can peer into other nodes to make sense of how they work

Scheduling automation (option: save and schedule as a job)

18 of 18

Come and visit us at Booth 905 !

www.sas.com/lifesciences

exscalate.com