Grand Ideas from CASPs:

1
(No Transcript)
2
Grand Ideas from CASPs
1. Computers help structure prediction no
more paper models
2. Knowledge-based potentials work better.
3. Local threading and fragment
assembly (Baker)
4. Averaging and consensus methods
work meta-servers (Ginalski-Rychlewski)
5. Sequence profile methods are as (or more
powerful) than threading (S?ding)
6. Jamming poorly similar templates
together helps (Skolnick-Zhang)
3
Main CASP8 results
1. no new grand ideas
2. not much progress in free modeling
3. some progress in moving away from templates
Tweakers vs. Transformers
Is it just engineering, or there is more to it?
Structure prediction it is not science, it
is engineering!
4. Mind over machine humans are still
surprisingly better
4
Date Mon, 2 Jun 2008 235639 -0500 (CDT) From
Nick Grishin ltgrishin_at_chop.swmed.edugt To David
Baker ltdabaker_at_u.washington.edugt Cc Ruslan
Sadreyev ltsadreyev_at_chop.swmed.edugt, Robert M
Vernon ltrvernon_at_u.washington.edugt Subject Re
C-terminus of T0407 I liked IG because of 1)
length 2) 7 strands 3) many IG are
interaction domains in enzymes. These
are very compelling reasons.
5
T0407_2 has Immunoglobulin fold
Cartoon diagram of 407, C-domain 3e38 chain A
residues 277-363
Cartoon diagram of VAP-A MSP Homology Domain 3z9l
6
No server predicted IG fold for T0407_2
Top GDT server model Phyre_de_novo TS1
T0407, C-domain 3e38 chain A residues 277-363
IG-based Baker model
7
http//prodata.swmed.edu/CASP8
8
Server rankings on all targets in domains for
three scores On 143 domains, ranking does not
change much with score, illustrating that the
ranking is robust.
9
Server rankings on FR domains for three
Z-scores On 28 FR domains, ranking shows small
variations illustrating the differences between
individual scores and between servers.  
10
Baker vs. servers energy spectra
FR mean of top 10 server GDT-TS between 30 and
52 - 21 domains
TR score (neutralizes compression effects),
First Z-scores
GDT-TS, Best Z-scores
GDT-TS, First Z-scores
http//prodata.swmed.edu/CASP8_secret/evaluation_s
ecret/DomainsAll.First.html
11
Baker vs. servers energy spectra
All targets (in domains) - 67 domains
TR score (neutralizes compression effects),
First Z-scores
GDT-TS, Best Z-scores
GDT-TS, First Z-scores
http//prodata.swmed.edu/CASP8_secret/evaluation_s
ecret/DomainsAll.First.html
12
ALL 67 human/server targets in domains for first
models, LGA GDT-TS
Sum of GDT-TS scores
Sum of Z-scores
13
ALL 67 human/server targets in domains for first
models, GDT-TS and TR scores
Sum of GDT-TS scores
Sum of TR-scores
14
ALL 67 human/server targets in domains for first
models, LGA GDT-TS
Sum of TR scores
Sum of Z-scores from TR
15
Why do we need human predictions?
The same reasons we need CASP.
If humans are better than servers we need
them. Analogy with chess.
16
My dream
Total server domination servers are so much
better than humans, so no human group can come
even close.
17
Acknowledgement
Our group
Collaborators
Shuoyong Shi Jing Tong Ruslan Sadreyev
Lisa Kinch Jimin Pei Ming Tang Sasha
Safronova Yuan Qi Hua Cheng
Jamie Wrabl Indraneel Majumdar Erik
Nelson Yong Wang S. Sri
Krishna Bong-Hyun Kim Dorothee Staber
David Baker and Co U. Washington Kimmen
Sjölander UC Berkeley William Noble
U. Washington
HHMI, NIH, UTSW, The Welch Foundation