Scribd
Upload
26 views1 page

Poster

1) The document presents a method for generating probability density functions (PDFs) from protein backbone dihedral angles to capture sequence-structure relationships. 2) It uses a self-consistent algorithm to construct PDF surfaces from Gaussian representations of dihedral angle distributions, which can then be converted to potential surfaces. 3) Tests on loop decoy sets and short protein loops show the native structures have among the highest probabilities compared to alternatives, supporting using PDFs to predict structure from sequence and sequence from structure.

Uploaded by

rataionel
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
26 views1 page

Poster

1) The document presents a method for generating probability density functions (PDFs) from protein backbone dihedral angles to capture sequence-structure relationships. 2) It uses a self-consistent algorithm to construct PDF surfaces from Gaussian representations of dihedral angle distributions, which can then be converted to potential surfaces. 3) Tests on loop decoy sets and short protein loops show the native structures have among the highest probabilities compared to alternatives, supporting using PDFs to predict structure from sequence and sequence from structure.

Uploaded by

rataionel
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 1

Statistical Potential for Backbone Sequence-Structure Interactions

Ionel Rata and Eric Jakobsson, UIUC


From Protein Coil Library data we create From the distribution plots we create dihedral Probability Density Functions (PDF) Ramachandran-like plots for adjacent backbone dihedrals in all possible residue contexts. PDF( i+1, i Asp-Asn ) Example: ( i+1, i) distribution for Ri = Asp and Ri+1= Asn
i i+1

Self-consistent convergence algorithm for PDF generation 1. Represent each ( n, a 2D Gaussian with
n) n

data point by = 20 degrees.

2. Construct a PDF surface by adding the Gaussians of all data points. 3.Compute new ns for each data point n ~ 1/sqrt(PDF( n, n)). 4. If any new n differs significantly from previous one GOTO step 2. 5. Else STOP. The PDF surface has converged.

that can be converted to potential surfaces using Boltzmann formula: -Potential = kTln(PDF)
i

i+1 i+1

Relative units

Nearest Neighbor Correlation (NNC)

Probability of a backbone structure given its sequence (structure prediction)


N i i

Probability of a sequence given its backbone structure (sequence design)

P(Str | Seq) ! PDF(


i !1
i+1 i+1 i

| R i -i R i R i i )NNC(

i 1

| R i R i1 ) Bayes formula:P(Seq | Str) ! P(Str | Seq) P(Seq)


P(Str)

Ri+1

Ri
i

N N -1 The native structure is among most probable P(Seq) ! P(R i-i R i R i i )/ P(R i R i  i ) from a set of alternative structures (decoys). i !1 i!1 We test this using the Loop Decoys Sets: The native sequence is among most probable

Loop Decoy Set 1NLS 1IXH 5PTI 1CSE 2PTH 1MSI 1FUS 5P21 1RCF_1 1RCF_2 2CTC_1 2CTC_2 1ABA 3SEB 1A2P 1A2Y 1EZM_1 1EZM_2

NNC(

i 1

| R i R i 1 ) !

PDF( i1 i | R i R ii ) PDF( i1 | R iR ii )PDF( i | R i R i i )

Calculating triple residue PDFs from single and double residue PDFs PDF( | R i -i R i R i  i ) ! PDF(
i i

| R i-1R i )PDF( i i | R i R i 1 ) PDF( i i | R i )

Structures Worse than Native 100% 98.5% 100% 98.9% 89.3% 100% 99.4% 93.7% 99.8% 97.1% 99.5% 100% 100% 99.3% 100% 98.7% 95.8% 100%

Loop Decoy Set 1MLA 1RIE 2ENG 1AKZ 153L 1A3C 1ADS 1ARU_1 1ARU_2 1BTK 1CVL 1DAD 1DIM_1 1DIM_2 1PPN_1 1PPN_2 3PTE

Structures Worse than Native 99.8% 99.7% 100% 100% 100% 100% 100% 100% 100% 99.9% 100% 99.7% 100% 99.4% 97% 100% 100%

from all possible sequences (20N). We test this using short loops from protein 1US0:
Loop Length Starting Resid Sequences Worse than Native 4 37 99.8% 4 109 99.1% 5 101 99.8% 5 151 99.999% 6 6 6 6 7 7 7 9 19 260 292 305 65 186 275 172 99.9% 99.9% 98.4% 99.9% 99.9% 99.998% 99.9% 99.9999%

You might also like