The Role of Openness in Scientific
Automation: a case for Open
Notebook Science
IJCAI'09 Workshop on Abductive and
Inductive Knowledge Development
Jean-Claude Bradley
Associate Professor of Chemistry
Drexel University
July 12, 2009

What is science?
FINDING OUT WHAT IS:
GOOD BAD
AND MAKING CHANGES:
BETTER CURE

Why are humans good at science?
GOSSIP CURIOSITY NEED TO ART
TO KNOW CREATE
UNDERSTANDING CONTROL
PREDICTIVE MODELS
CONCEPTUAL (docking) EMPIRICAL (QSAR)

The standard model of how science
happens
Propose Design
hypothesis experiments
Analyze Perform
results experiments

The actual way much of science
happens
Think up Find collaborators Convince
something who also think what somebody to
cool/useful to you want to do is fund your
do cool cool idea
Find existing Do the work Share what you
data or (experiments) found with the
models to
world
minimize the
amount of
Usually last (and
trial and error
only a small part)

How can the scientific process become
more automated?
WE ARE HERE

The Robot Scientist

How radical openness can assist in the
automation of science
 Agents can participate with zero or near-
zero cost (free hosted services – e.g.
Google)
 Self-organizing redundant processes

How can machines know what is
important?
Ask the humans

UsefulChem Blog

What chemists think is important in
2005

Find-A-Drug

Open and Closed Science
Traditional Traditional Open Notebook
Lab Notebook Journal Open Access Science (full
(unpublished) Article Journal Article transparency)
RESEARCH
CLOSED OPEN
TEACHING
Lectures Archived
Traditional Assigned
Notes Lectures
Paper problems
public Public and
Textbook public free online
F2F lectures
textbooks

Motivation: Faster Science, Better Science

Open Notebook Science Logos
(Andy Lang, Shirley Wu)
Sharing: how much and when

There are NO FACTS,
only measurements embedded
within assumptions
Open Notebook Science maintains
the integrity of data provenance by
making assumptions explicit

TRUST
PROOF

The solubility of 4-chlorobenzaldehyde

The Log makes Assumptions Explicit

The Rationale of Findings Explicit

Raw Data Made Public
Splatter?
Some liquid

YouTube for demonstrating experimental set-up

Calculations Made Public on
Google Spreadsheets

Revision History on
Google Spreadsheets

Wiki Page History

Comparing Wiki Page Versions

Proof of Purity with interactive NMR
spectrum using JSpecView and JCAMP-DX

Linking to Molecules in Chemistry
Databases

Experimental Spectra and User-
Deposited Data on ChemSpider

Open Data JCAMP spectra for education
(Andy Lang, Tony Williams)
(Andy Lang, Tony Williams, Robert Lancashire)

Database Curation via Game Playing

Over 50,000 spectrum views so far -
worldwide

Link Spectral Game to Open Educational
Content

NMR game in Second Life (Andy Lang)

Crowdsourcing Solubility Data

ONS Challenge Students

ONS Challenge Judges

Teaching Lab: Brent Friesen
(Dominican University)

Solubility Experiment List

Solubilities collected in a Google
Spreadsheet

Rajarshi Guha’s Live Web Query
using Google Viz API

Rajarshi Guha and Andy Lang:
Chemical Space Explorer

Semi-Automated Measurement of
solubility via web service analysis
of JCAMP-DX files
(Andy Lang)

Solubility Measurement Requests:
DoSol sheet
•Outlier Bot: flags measurements with high standard deviation to mean
ratios
•Google Analytics queries – new solvent/solute searches
•Solubility request form – researcher in Israel requesting pyrene in
acetonitrile solubility for environmental soil contamination study
•Application based models – high priority Ugi reactants

Solvent mixture and temperature:
multidimensional solubility data
Actual Data From quadratic regression
(4-nitrobenzaldehyde) of 5D space
Feeds DoSol Sheet the next points to measure to best cover the space

Crowdsourcing Reaction Requests:
DoUgi sheet

Understanding in addition to empirical
modeling
Missed in a prior
publication on solubility
for this compound

Data provenance:
From Wikipedia to…

…the lab notebook and raw data

Including links to the literature

Pierre Lindenbaum’s Solubility Data
as RDF Triples

Results and Workflows in Machine-
Friendly Format

Experiments in Chemical Markup Language

UsefulChem Project: Open Primary
Research in Drug Design using Web2.0
tools
Rajarshi Guha Tsu-Soo Tan
Indiana U Docking Nanyang Inst.
JC Bradley
Drexel U Synthesis
Phil Rosenthal Dan Zaharevitz
UCSF Testing NCI
(malaria) (tumors)

Malaria Target: falcipain-2
involved in hemoglobin metabolism
Dana.org

Where’s the Beef?

Link to Lab Notebook Page in Wiki

Link to Docking Procedure
(Rajarshi Guha)

Link to Docking Results:
Lists of SMILES in GoogleDocs
(Rajarshi Guha)

Outcome of Guha-Bradley-Rosenthal
collboration

Collaborative Drug Discovery (CDD) Database

How does Open Notebook Science fit with
traditional publication?
•Concentration (0.4, 0.2, 0.07 M)
•Solvent (methanol, ethanol, acetonitrile, THF)
•Excess of some reagents (1.2 eq.)

Mettler-Toledo MiniMapper

Mettler-Toledo MiniBlock System

XML reports from MiniMapper robot

GoogleDoc to program and report

Paper written on Wiki

References to papers, blog posts,
lab notebook pages, raw data

Paper on Journal of Visualized
Experiments (JoVE)

Pre-print on Nature Precedings

ChemSpider Automated Mark-up of
Chemical Names

Archiving and Curation of ONS and
other new forms of scholarship?

Other Open Notebooks
Cameron Neylon’s Notebooks

Anthony Salvagno’s Notebook
(Steve Koch group)

Acknowledgements
 Khalid Mirza (Drexel)
 Jenny Hale (Southampton U.)
 David Bulger (Oral Roberts U.)
 Tim Bohinsky (Drexel)
 Kevin Owens (Drexel)
 Tom Osborne (Mettler-Toledo)
 Antony Williams (ChemSpider)
 Andrew Lang (Oral Roberts U.)
 Rajarshi Guha (Indiana U.)
 Cameron Neylon (Southampton U.)