1. The Semantics of Drug Discovery and Marketing:
How to Empower Knowledge Workers and Save
Money
Rob Gonzalez
Senior Product Manager
Cambridge Semantics Inc.
rob@cambridgesemantics.com
10. 11
The Easy Problems have been Solved
The Environment is Becoming more
Complex
New Technologies are being Introduced at
an Increasing Rate
There is a Deluge of Data
14. 15
"No plan of battle ever survives contact with the enemy.”
- Heinz Guderian
“The only constant is change.”
-Herakleitos of Ephesus, cerca 500 B.C.E.
“This too shall pass.”
-Proverb
Traditional IT
15. Flexibility in the face of Changing Requirements
In business, new situations arise that require
previously unknown data from unanticipated
sources presented in unexpected ways to end
users.
16. Case Studies:
1. Assay Management in Drug Research
2. Manufacturing Quality Control
3. Salesforce Optimization
25. 26
Anzo Data Collaboration
Server
Anzo Solution
Provides Scientists:
• Sub-hour turnaround
for root cause analysis
(vs. WEEKS before!)
• Familiar day-to-day
tools (Excel for data
entry, Web for
analysis)
Provides IT:
• Rapid Solution to
Integration Problem (1
week initial
implementation)
• Full audit trail for
compliance
• Low Capex
• Access control and
governance over
sensitive data
26. Linking is a Semantic Mapping of
Spreadsheet Data to a standard model
28. Pharma Sales is Collaborative
Situation:
•Market position of leading drug
is threatened by new entrants
•Each of 100+ reps has hundreds
of Drs. in his territory
•Whom should each speak with
today? This week? This month?
•Heuristic requires two-way data
communication. Data in heuristic
likely to change
•Existing data collaboration was
done via spreadsheets
30. Pharma Sales is Collaborative
Anzo Data Collaboration
Server
Provides Reps:
• Optimal lead identification
• Familiar tools with zero
training required
Provides Sales Management:
• Ability to protect holistic
market position
• Ability to go on offensive in
specific regions
• More accurate sales
forecasting (the reps actually
use the CRM!)
• Ability to change heuristic
without requiring IT
intervention
31. 32
The Easy Problems have been Solved
The Environment is Becoming more
Complex
New Technologies are being Introduced at
an Increasing Rate
There is a Deluge of Data
32. Semantic Technologies Handle this through:
– Flexibility in the Face of Inevitable Change
– Focusing on End Users
– Providing a Rapid Time to Market for Solutions
33. • Salesforce Optimization
• Departmental Budget Management
• Drug Discovery and Basic Research Data Collection & Reporting
• Manufacturing Quality Control
• Data Collection and Reporting for Scouts in the Field
• Clinical Trial Data Analytics
• …
Cambridge Semantics Solutions in Life Sciences
I’m here to tell you how Semantic Technologies can help stop some of the cost hemorrhaging of the drug industry, by helping knowledge workers, including drug researchers, manufacturers, and marketers use data they are already collecting to make better decisions.
But first I’d like to give you some context on the drug industry and the challenges that it faces today.
Flashback to 1945. Here we see Sir Alexander Fleming being awarded the Nobel Prize for his discovery of Penicillin, the drug that changed the world. No invention by mankind has been responsible for saving so many lives. No drug discovery more significant.
And he did it by accident. He literally left a Petri dish out for too long and noticed that the resulting fungus killed the bacteria he was studying! The greatest medical discovery of all time, and it was by mistake, in a little dish that today costs less than a dollar.
Sir Alexander Fleming (6 August, 1881 – 11 March, 1955)
Image source: Wikipedia
http://en.wikipedia.org/wiki/Alexander_Fleming
Fast forward to today’s world. If you’ve been following the news, the strains of antibiotic resistant bacteria are growing in number and variety. There is a strain of TB that is resistant to all but one antibacterial drug, and is a step away from becoming a major epidemic.
Chart Source: The Acute Respiratory Infections Atlas
http://www.ariatlas.org/prevention_diagnosis_treatment/antibiotics_and_antiviral_therapy
At the same time, the number of new antibiotic treatments hitting the market is declining with each passing year.
In fact, in 2002, out of 89 new drugs, no new antibiotics were approved.
Chart Source: The Acute Respiratory Infections Atlas
http://www.ariatlas.org/prevention_diagnosis_treatment/antibiotics_and_antiviral_therapy
Quote Source: Infectious Diseases Society of America
http://www.idsociety.org/Content.aspx?id=5558
To make an analogy; at the beginning finding the drugs was easy. Sir Alexander Fleming tripped over the greatest drug mankind has ever known.
However, the obvious discoveries have all been made. With each passing year, the bar is set higher to finding new cures to diseases.
What’s more, not only are new drugs harder to identify, but we’re also much more careful about how fast we unleash them on the general population, and for good reason. Our new fascination with the microbial world caused significant damages.
One very notable example is Thalidomide, which was introduced in the 1950s in Europe and caused significant birth defects in children.
This images is of the DA inspector Frances Oldham Kelsey receiving an award from President John F. Kennedy for blocking sale of Thalidomide in the United States.
Image source: Wikipedia
http://en.wikipedia.org/wiki/Thalidomide
This story is not all bad news, however. As our knowledge of the human body has increased, so has our ability to dream up new cures. Genetic therapies are being researched.
New, automated compound testing facilities can screen thousands of varieties of compounds against thousands of diseases in a highly automated way searching for new potential cures.
Image source: Fluofarma
http://www.fluofarma.com/Toxicology/predictive-toxicity-assays.html
Even robots—which are, as yet, non-threatening to us—have found a way to pitch in.
To sum up:
The Easy Problems are Solved
The Environment is Becoming more Complicated
New Technologies are Introduced at an Increasing Rate
There is a Deluge (Bounty!) of Data
But there is a cost to working in this world…
But with new complexity comes rising research costs, which you have also no doubt heard in the media. We are literally inventing the technology used to research drugs as we’re doing it, which is hard.
In fact, R&D expenses have grown 7.4% faster than inflation, with costs continuing to rise.
Chart Source: The price of innovation: new estimates of drug development costs. DiMasi, et al., 2002.
This is true in every phase of development, from candidate compound identification in a lab, through clinical trials, manufacturing, and market research.
Traditional computing, embodied here by a process model for creating a data warehouse, is very rigid. The first thing you have to do before building a large IT solution is to gather requirements. Then you have to build a logical data model. Then you build a test system. Etc.
This process, while effective, takes a long time—which is the whole point. By the time a new solution is introduced, the world has already moved on. For example, it took hospitals more than 6 months after the outbreak of SARS before their computer systems even had a code to track its occurrence. The world simply does not stand still in accordance with our well thought out data models.
As Lee said in his opening, and as others have said throughout this show already, a key strength of the semantic technology stack is that it is inherently flexible. It is made to change with requirements, and change with your business.
As promised, I am going to tell you about 2 examples of semantic solutions in pharmaceutical companies that save money.
As I was thinking about which to talk about I decided that I didn’t have the time necessary to go into the details of scientific drug discovery, which is pretty complicated. But fortunately for you, it has been presented by Merck in the past at SemTech. If you want specifics, come find me after.
The first is a story of a couple applications we build around manufacturing quality control for a major drug manufacturer, aimed at bringing costs down.
The second is an application that we built for a biotech company to help their salesforce drive revenues up.
This is a basic overview of how things are generally manufactured in the pharma world today, which is shockingly similar to how we make anything else. Personally, I would have expected drug manufacturing to be somehow flawless, but the reality is that it’s not! They outsource the production of some compounds overseas, ship stuff around, etc.
To compensate, there are strict procedures for quality control, including a necessary post-mortem. Making drugs is not like making toys; it is a highly regulated industry, and there has to be root cause analysis whenever anything goes wrong.
There are many compounds and processes involved in drug manufacturing. Each one has a great deal of data. Data is also collected about the manufacturing process itself, as well as about the shipping process.
All of this information is stored in Excel spreadsheets in a document management system.
But why not build a standard IT system to solve the problem? Why take a risk on Semantic Technology?
Well, the first reason is simply cost. They looked into a traditional IT solution to the problem, and it would have taken at least a year and cost much more than they were willing to spend.
The system is not flexible enough to handle this environment without ridiculous amounts of back end work.
One way to get flexibility is to use spreadsheets. This is a major reason that they were using spreadsheets, and why many businesses use spreadsheets for vital functions instead of hardened IT systems. A spreadsheet can evolve to contain new kinds of information very quickly, and can be used directly by end users.
At Cambridge Semantics we offer a Semantic Middleware Platform that provides the scalability and enterprise data features required to tackle today’s business problems. The solution to the manufacturing problem is built on our middleware platform, and took full advantage of the unique Excel plugin that we provide.
Our solution required each spreadsheet to the linked to the data collaboration server. Once that linking happened, scientists could perform root cause analysis in a familiar environment that did not require retraining, and a ton of time has been saved.
The linking process is a semantic mapping of spreadsheet data to a standard model. It is an easy process that takes minutes to do for most standard spreadsheets. In fact, Lee Feigenbaum built an entire application from scratch in front of a live audience at SemTech in June.
I’m not going to demo the product here, but definitely get in touch if you want to see it work in depth.
The situation facing the drug company was not an enviable one. The market position for their blockbuster drug was being eroded by new entrants, and they needed new strategies to protect their existing market and regain lost ground.
Their direct connection to Drs. Prescribing the use of their treatment is, of course, their salesforce. Each rep in the force has a territory consisting of hundreds of Drs., and the company had developed a heuristic to determine which Drs. Each should focus on.
The heuristic required data flow in both directions and might CHANGE over time.
Oh, and, like everyone else, they use spreadsheets.
I can’t give away the details of the heuristic, nor can I show you the exact spreadsheets used or all the data required in the solution.
I can, however, illustrate one very important detail.
There are two specific kinds of spreadsheets needed in the collaboration. One contained master sales data spanning many reps over time to feed some parts of the heuristic. The other contains data from individual reps, including this quarter’s sales estimate, which is a bit of a guess and can’t be automated.
Thus BOTH human input AND calculated data is required to make this work.
To sum up, here is the state of the world that we presented early in the talk.
We then showed you two real-world solutions in which semantic technology has played a major part.
Now, think about the analogies in your own industries, with Oil & Gas (used to need a hardy straw and a set of lungs, now require 30,000 ft. of tubes and an off shore drill), or cooking (cheese was discovered like penicillin, but now we’re doing molecular gastronomy) , or advertising (microsecond ad auctions), or finance (as you heard in yesterday’s panel).
I talked about two solutions today, but Cambridge Semantics has built many more in Life Sciences alone.
We have a number of marquee customers and partners, both within and outside of Life Sciences.
Finally, I didn’t want to be left out, so I included the obligatory Semantic Web slides every deck must have.
…and the layer cake.
I hope you enjoyed this talk, and look forward to your questions.