2. Executive Summary
+ Tomorrow’s challenge is to develop new medicines that can
prevent or cure currently incurable diseases
+ Refresh the pharmaceutical drug pipeline, by taking better
advantage of the available data with new algorithms and
disruptive technologies
+ Artificial Intelligence (AI) in Pharma R&D will help to identify
and validate new drug targets, support early identification
of safety and efficacy issues, and improve patient
stratification
6. Breast Cancer Diagnoses - 2017
Pathologist Performance A.I. Performance
https://research.googleblog.com/2017/03/assisting-pathologists-in-detecting.html
73% 92%
Doctors often use additional tests to find or diagnose breast cancer
The pathologist ended up
spending 30 hours on this
task on 130 slides
A closeup of a lymph node biopsy.
7. Skin Cancer Diagnoses - 2016
Pathologist Performance A.I. Performance
http://www.nature.com/nature/journal/v542/n7639/full/nature21056.html
96,5% 97,1%
If found early 95% of skin cancers can be treated successfully
Pathologist + A.I.
99,5%
8. Early Diagnosis of Congestive Heart Failure
http://ml.gatech.edu/
A machine learning
example from Georgia
Tech demonstrated that
machine-learning
algorithms could look at
many more factors in
patients’ charts than
doctors, and by adding
additional features there
was a substantial
increase in the ability of
the model to distinguish
people who have CHF
from people who don’t.
Human
performance
A.I.
performance
9. Predict Cardiac Failure Before It’s Diagnosed
https://arxiv.org/abs/1602.03686
In quantitative evaluation, our proposed representation significantly improves
the predictive modeling performance for onset of heart failure (HF), where
classification methods achieve up to 23% improvement in area under the
ROC curve (AUC) using this proposed representation.
12. Distributed Machine Learning in Data Center
Data Size
Model Size
Model parallelism
Single machine
Data center
Data
parallelism
training very large models exploring several model
architectures, hyper-
parameter optimization,
training several
independent models
speeds up the training
13. Machine Learning Workflow
Collect data
Data
Preprocessing
Search
Analysis
Model
Training
Re-
simulation
Reports
Results
Model
Deployment
Training
data
Model
Testing
Train Test Loop
Test
data
Model Feedback Loop
14. Think Big Business Strategy
Data Strategy
Technology Strategy
Agile Delivery Model
Business Case Validation
Prototypes, MVPs
Data Exploration
Data AcquisitionStart Small
Value
Proposition