A World of New Discovery

1
A World of New Discovery
STRUCTFAST Novel Dynamic Programming and Profile
Scoring CASP6 CAFASP4 Servers EXPM, BNMX
SFST Aleksandar Poleksic, Joseph F. Danzer,
Derek A. Debe
2
Derek A. Debe, Aleksandar Poleksic, Joseph F.
Danzer, respectively.
3
OutlineWhat's Different About Our Methods?
1. Bridge-Bulge Dynamic Programming 2.
Analytical Profile-Profile Scoring 3. Convergent
Island Statistics p-Values
4
STRUCTFAST
Structure Realization Utilizing Cogent Tips From
Aligned Structural Templates
Basic Principle Gaps known to exist should not
be strongly penalized.
Known Gap
Known Gap
Structure Alignment of Homologous Crystal
Structures
Structure-Structure alignments give us clues...
5
Bridge-BulgeDynamic Programming
Dynamic Programming
STRUCTFAST
BIG-TOWNSOWN--- BIGBROWNTOWNOWN
BIG-----TOWNSOWN BIGBROWNTOWN-OWN
6
CASP Target T0247
Template 1pj5A. Bridge taken from
1wooA Query 2 AQ---QTPLYEQHTLCGARMVDFHGWMMPLH
YGS-------------------------- 32
P H GA GW P Sbjct 427
PrnlrVSPFHARHKELGAFFLEAGGWERPYWFEAnaallkempaewlppa
rdawsgmfss 486 Query 33 --QIDEHHAVRTDAGMFDVS
HMTIVDLRGSRTREFLRYLLANDVAKLtkSGKALYSGMLN...
E RT M D G L L
D AK G Y L Sbjct 487
piAAAEAWKTRTAVAMYDMTPLKRLEVSGPGALKLLQELTTADLAKKPG
AVTYTLLLD... Query 91 ASGGVIDDLIVYYFTEDFFRLVV
NSATREKDLSWITQHAEPF-----GIEI-TVRDDLSM 144
GGV D V ED F L N
H Sbjct 545 HAGGVRSDITVARLSEDTFQLGANGNIDTAYFERA
ARHQTQSgsatdWVQVrDTTGGTCC 604 Query 145
IAVQGPNAQAKAATLFN-DAQRQAV-EGMKPFFGVQAGDLFIATTGYTGE
AGYEIALPNE 202 I GP A D
Y GE G E Sbjct 605
IGLWGPLARDLVSKVSDdDFTNDGLkYFRAKNVVIGGIPVTAMRLSYVGE
LGWELYTSAD 664 Query 203 KAADFWRALVEAG----VKP
CGLGARDTLRLEAGMNLYGQEMDETISPLAANMGWTIAWE 258
W AL AG V G A LRLE G G
M P A G Sbjct 665 NGQRLWDALWQAGqpfgVIAAGR
AAFSSLRLEKGYRSWGTDMTTEHDPFEAGLGFAVKMA
724 Query 259 paDRDFIGREALEVQREHG-TEKLVGLVMTE-
KGVLRNELPVRFtdaqGNQHEGIITSGT 316
FIG ALE E L L PV Q
G TS Sbjct 725 --KESFIGKGALEGRTEEAsARRLRCLTIDD
gRSIVLGKEPVFY----KEQAVGYVTSAA 778 Query 317
FSPTLGYSIALARVPEGI--GETAIVQIRNREMPVKVTKPVFV-RNGKAV
363 T IA P G
R VT Sbjct 779 YGYTVAKPIAYSYLPGTVsvGDSVDIEY
FGRRITATVTEDPLYdPKMTRL 828
1wooA 1pj5A 1pj5A Bulge
7
CASP Target T0263
Template 1iujB. Bridge taken from 1a7gA
Query 1 MaTLLQLHFAFN-GpFGDAMAEQLKPLAESINQEPGF
LWKVWTESEknhEAGGIYL---- 55 M
E A PGF G
Y Sbjct 1 M-FVTMNRIPVRpE-YAEQFEEAFRQRARLVDRMP
GFIRNLVLRPK---NPGDPYVvmtl 55 Query 56
FTDEKSALAYLE--KHTARLKNLG-------VEEVVAKVFDV 88
E A E G
F V Sbjct 56 WESEEAFRAWTEspAFKEGHARSGtlpkeafLG
PNRLEAFEV 97
1a7gA 1iujB 1iujB Bulge
8
OutlineWhat's Different About Our Methods?
1. Bridge-Bulge Dynamic Programming 2.
Analytical Profile-Profile Scoring 3. Convergent
Island Statistics p-Values
9
STRUCTFAST Scoring

Score ?

Seq1 VLKRAKAKGDRLILEADGEKNKLNVIF
Seq2 FRGVTLSAKEFKKIVDTFSQLSDSVNFQVDKEGIKL
L D B

Profile-Profile Case
10
STRUCTFAST Scoring
Like Throwing Dice
I
D
L
K
T
A
W
L
K
H
Which die is more likely to generate D?
K

D
B
N
G
S
R
R
M
D
C
V
A


Score log(PL(D) / PB(D))

We solve this analytically. Details to be
published...
11
How do we evaluate the statistical significance
of our STRUCTFAST alignments?
12
The Challenge
Secondary Structure
Profiles
Parameters (Gaps, et al.)
Bridges Bulges
Rigorous Statistical Significance of the
Alignment (p-Value)
13
Island Statistics Altschul, et al.
Not fast enough for full database search
14
.










.







.










.








Convergence Trick

K
Higher p-values
Lower p-values
Lower p-values
Higher p-values
True
True K
We shuffle until there is evidence that
the sequences are not related.
15
Benefits of CIS (Convergent Island Statistics)
Analytical e-values No need for distribution
fitting Analytical (yet tunable) trade-off
between speed and sensitivity Allows one to
change any algorithm parameters on the fly