Download presentation
Presentation is loading. Please wait.
Published byAmberlynn York Modified over 9 years ago
1
T HE S CREENERS WERE CREATED BY MAN
2
T HEY EVOLVED
3
A ND THEY HAVE A PLAN
4
Drug Screening Test lots of compounds for inhibition against a target
5
High Throughput Screening
6
What if you do not know the target?
7
Not all targets are known 1. Compound X extracted from Australian Jumping Fungus kills cancer cells. What is its mechanism?
8
Not all targets are known 1. Compound X extracted from Australian Jumping Fungus kills cancer cells. What is its mechanism? 2. A collection of leads has been designed for a particular target, do they have any off-target effects?
9
High Content Screening (image from http://www.encorbio.com/monoclonal/MCA-39C7.htm)
12
Questions 1. Can we profile the biological activities of a library of compounds efficiently? 2. Can we use simple probes of the cell cycle to obtain high-resolution biological information? 3. How do we draw conclusions from the massive amounts of data generated?
13
The Screening Library 6,547 Compounds 58% natural products 21% known bioactives 21% structural diversity
14
The Setup Add Hela cells to wells in 384 well plates Add 20 μM compound Incubate 20 hrs Fix, stain
15
The Fluorescent Probes Hoescht 33342 Labels DNA (DAPI) G1/G2 Immunostain Labels Histone H3 Phospho-Ser10 Mitotis Click Chemistry Labeled nucleotide pulse DNA Replication
16
High-throughput Image Analysis #FeatureDescriptionValue 1AreaCh1Nuclear Area X 1 2PerimCh1Nuclear Perimeter X 2 36 IntenCoocASMCh3 EdU Texture X 36 Image 500 cells/well Analyze 36 features from 3-channel fluorescence images
17
Factor Analysis Assume measured variables are linear functions of common, underlying factors. Reduce dimensionality of data to factors that explain common sample variance
18
Factors represent meaningful phenotypes
19
Compounds can be ranked by activity For each factor: (treated – untreated)^2
20
Compounds can be ranked by activity For each factor: (treated – untreated)^2 Six Factors
21
Compounds can be ranked by activity For each factor: (treated – untreated)^2 Six Factors A.J. Hanson/Indiana Univ.
22
Compounds can be ranked by activity For each factor: (treated – untreated)^2 Six Factors
23
Compounds can be clustered by factor scores
24
Compounds cluster into distinct phenotypic groups
25
Phenotypic similarity correlates with structural similarity Phenotype vector Euclidian Distance 2D Fingerprint (ECFP_4) Tanimoto coefficient
26
Structural similarity does not always correlate with phenotypic similarity
27
A: D2 receptor antagonist B: Binds many GPCRs including D2 dopamine receptor
28
Compounds cluster into distinct phenotypic groups
29
Phenotypic similarity captures fine details of activity
30
Cytotoxic macrocylic hexadepsipeptides
31
Phenotypic similarity captures fine details of activity Cytotoxic macrocylic hexadepsipeptides Cytotoxic macrocyclic nonpeptides
32
Clusters can represent mechanisms of action G1/G0 Arrest Phenotype
33
Clusters can represent mechanisms of action Cardiac glycosides inhibit Na/K pumps G1/G0 Arrest Phenotype Can inhibit translation at high levels
34
Clusters can represent mechanisms of action G1/G0 Arrest Phenotype Steroid Hormones
35
Clusters can represent mechanisms of action Cardiac glycosides inhibit Na/K pumps G1/G0 Arrest Phenotype Can inhibit translation at high levels
36
Clusters can represent mechanisms of action Cardiac glycosides inhibit Na/K pumps G1/G0 Arrest Phenotype Can inhibit translation at high levels Translation inhibitor
37
Conclusions Factor analysis can be applied to reduce a huge amount of microscopy data to interpretable biological phenotypes A small number of probes is capable of detecting fine distinctions in compound activity
38
Caveats Could not connect compounds to individual targets
39
Caveats Could not connect compounds to individual targets Is there even a single target?
40
Acknowledgements Faculty Coaches Brian Shoichet Jack Taunton Jim Wells Student Coaches Noah Ollikainen David Booth Heather Newman “One useless compound is called a placebo, two useless compounds are called a controlled experiment, and three or more become a screening library!”
41
Acknowledgements Faculty Coaches Brian Shoichet Jack Taunton Jim Wells Student Coaches Noah Ollikainen David Booth Heather Newman “One useless compound is called a placebo, two useless compounds are called a controlled experiment, and three or more become a screening library!”
43
All Image Analysis Variables
44
384-well Optical Plates 2000 HeLa Cells/well Grow overnight Add 20 μM compound Incubate 37 o 20hr EdU pulse, Fix The Setup (image from http://www.encorbio.com/monoclonal/MCA-39C7.htm)
45
Factor Analysis Assume measured variables are linear functions of common, underlying factors. Select factors that explain common variance
46
Different compounds vary in potency Phenotypic Response Metric Distance in factor space from control, untreated cells Select “hits” as compounds with top 5% phenotypic response in both replicates 211 compounds (3%)
47
Phenotype-structure correlation is statistically significant
49
Clusters can represent mechanisms of action Cardiac glycosides inhibit Na/K pumps G1/G0 Arrest Phenotype Can inhibit translation at high levels Translation Inhibitor Another translation inhibitor!
50
“One useless compound is called a placebo, two useless compounds are called a controlled experiment, and three or more become a screening library!” High Content Screening and You
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.