crick’s early hypothesis revisited

47
Crick’s early Hypothesis Revisited

Upload: alia

Post on 05-Jan-2016

53 views

Category:

Documents


2 download

DESCRIPTION

Crick’s early Hypothesis Revisited. Or The Existence of a Universal Coding Frame Axel Bernal UPenn Center for Bioinformatics Jean-Louis Lassez Coastal Carolina University Ryan Ross Coastal Carolina University. BIOINFORMATICS. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Crick’s early Hypothesis  Revisited

Crick’s early Hypothesis Revisited

Page 2: Crick’s early Hypothesis  Revisited

Or The Existence of a Universal Coding Frame

Axel BernalUPenn Center for Bioinformatics

Jean-Louis LassezCoastal Carolina University

Ryan RossCoastal Carolina University

Page 3: Crick’s early Hypothesis  Revisited

BIOINFORMATICS

The application of computer technology to the management and analysis of biological data

COMPUTATIONALBIOLOGY

Page 4: Crick’s early Hypothesis  Revisited

Biology: the study of living organisms

Why should computer scientists be interested in biology?

Page 5: Crick’s early Hypothesis  Revisited
Page 6: Crick’s early Hypothesis  Revisited

Genomes and GenesThe language of life

…..catgcctagactgcatcggtaccatgacatgcatttatagaacactacgcgtaatagccatgatcccatagatacatacagagatacactgatagactcgacctcatccgattatatagacctgaaatggctagctggacatgcgatcgaatcgagattagcaccatagagtggcatagccatgcgctgatagcaaaatgccatagctagtgtctaacgtgcattgccctggatgacatggctccgatatggcggctgatcgtcgctgaaatgctcgctgcaatggctaggatacagtaatagacgtaatgccaatggctgctcgctggatagtcgctgacatcgatcgcctgatatgatgcgctagctccgcataagatcgctgatcgcta……..

Page 7: Crick’s early Hypothesis  Revisited

Genetic Code

Page 8: Crick’s early Hypothesis  Revisited

Crick’s 1957 Hypothesis

The genetic code has excellent information theoretic properties, it is

comma free

It does not admit ANY form of parasitism.

Page 9: Crick’s early Hypothesis  Revisited

Dismissed for the past 35 years

Replaced by “Frozen Accident”

• Renewed interest in comma free and circular codes (DNA computing, Arques/Michel)

• Time to revisit

Page 10: Crick’s early Hypothesis  Revisited

Coding

0000 = A1111 = B0001 = C1000 = D0011 = E1100 = F0111 = G1110 = H

0010 = I0100 = J0101 = K1010 = L1001 = M0110 = N1011 = O1101 = P

Page 11: Crick’s early Hypothesis  Revisited

0010111010110010111100011011100011011001110101010010101100110111

Communication Error

0010111010110010111100011011100011011001110101010010101110110111

I H O I B C O D P M P K I O E G

I H O

X

K H E G C O E L L K G N ...

Translation ErrorFrameshift

Page 12: Crick’s early Hypothesis  Revisited

…101011100100111010010010001010111…

Parasite sub Messages

Bounded Parasitism:

…101011100100111010010010001010111…

Spread Parasitism:

Page 13: Crick’s early Hypothesis  Revisited

Biological Implications of comma free

A frameshift will immediately abort the translation

ANY fragment of length 5 in the coding region of ANY gene in ANY organism determines the frame

Universal Frame property

Page 14: Crick’s early Hypothesis  Revisited

Crick’s Hypothesis Revisited

What is the length of the shortest segment of a coding region that defines the frame independently of the organism it comes from?

IF IT EXISTS

Page 15: Crick’s early Hypothesis  Revisited

Mathematical Concepts

Comma Free Codes

Codes with Bounded and Spread Parasitism

Circular codes

Locally Testable Languages

Similarity Measures

Page 16: Crick’s early Hypothesis  Revisited

A Circular Code

1

01

001

0001

00001

000001

0000001

Page 17: Crick’s early Hypothesis  Revisited

Unique Decomposition

Page 18: Crick’s early Hypothesis  Revisited

A Non Circular Code

000111 001 100011 110 101010

Page 19: Crick’s early Hypothesis  Revisited

Multiple Possible Decompositions

Page 20: Crick’s early Hypothesis  Revisited

00101110101100101111000110111000110101111101010100101011101101110010111010110010111100011011100011010111110101010010101110110111

0011111011110010011100011011100011011001110001110110100110110111

Locally Testable Events

∑* / ∑* 0101 ∑*

Page 21: Crick’s early Hypothesis  Revisited
Page 22: Crick’s early Hypothesis  Revisited
Page 23: Crick’s early Hypothesis  Revisited
Page 24: Crick’s early Hypothesis  Revisited

Theorem

Assumption: code X consists of a finite set of words all of the same length

The following are equivalent:

X has bounded parasitism of degree dXd+1 is comma freeX is circularX* is strictly locally testable

Page 25: Crick’s early Hypothesis  Revisited

Crick’s Hypothesis Revisited Again

Genetic code C

Language of Genes G≠C*

C has good properties then G has good propertiesBUT G may have good properties while C does not.

Shift from comma free to Testable by fragments

Page 26: Crick’s early Hypothesis  Revisited

Similarity

CX

uC

u

XXSXS ),()(

YXS ,

)(maxarg)( XScX C

2

2

2

YX

e

Page 27: Crick’s early Hypothesis  Revisited
Page 28: Crick’s early Hypothesis  Revisited
Page 29: Crick’s early Hypothesis  Revisited
Page 30: Crick’s early Hypothesis  Revisited

Arques/Michel Codes 1998

00 },{ XTTTAAA 11 }{ XCCC 22 }{ XGGG

X0 = {AAC, AAT, ACC, ATC, ATT, CAG, CTC, CTG, GAA, GAC, GAG, GAT, GCC, GGC, GGT, GTA, GTC, GTT, TAC, TTC}

X1 = {ACA, ATA, CCA, TCA, TTA, AGC, TCC, TGC, AAG, ACG, AGG, ATG, CCG, GCG,

GTG, TAG, TCG, TTG, ACT, TCT}

X2 = {CAA, TAA, CAC, CAT, TAT, GCA, CCT, GCT, AGA, CGA, GGA, TGA, CGC, CGG,

TGG, AGT, CGT, TGT, CTA, CTT}

Page 31: Crick’s early Hypothesis  Revisited

T Representations

Frame0:Frame1:Frame2:

ATGGGCAAGTAA

1 0 1 22 222 2 0

ATGGGCAAGTAAATGGGCAAGTAA

Page 32: Crick’s early Hypothesis  Revisited

Training set

• DKEYP-117 zebra fish gene.

• KEGG

• 10620 Nucleotides

• Length of windows 200 in T representation

• C is 1671 Windows (Coding frame)

• C++ 1670 Windows

Page 33: Crick’s early Hypothesis  Revisited

First Experiment

Page 34: Crick’s early Hypothesis  Revisited

• Consistent with Crick’s hypothesis but for the size of the code.

• Comma-free code (words of length 600)

OR

• G is locally testable

• Robustness with respect to overfitting.

Page 35: Crick’s early Hypothesis  Revisited

General ExperimentData sets

• We selected 14 different organisms in all three families and extracted 50 genes from each (Ecoli, Pyrococcus, Anopheles gambiae….).

• 100 genes which were selected from KEGG, NCBI, Weizmann Institute (TP53, Atm, HIV, Breast cancer…).

• 1000 genes with various ranges of GC Contents (Center for Bioinformatics, UPenn).

Page 36: Crick’s early Hypothesis  Revisited
Page 37: Crick’s early Hypothesis  Revisited

• Not Comma-free• Maybe Bounded Parasitism/Circular• It is testable by fragments

ATG…GGCAA…CACC…TAATGA..AGTG…CCAA..ACCCT…GCAAC..TAG…

Page 38: Crick’s early Hypothesis  Revisited
Page 39: Crick’s early Hypothesis  Revisited

ATG…GGCAA…CACC…TAATGA..AGTG…CCAA..ACCCT…GCAAC..TAG…….

• Not Comma-free• Not Bounded Parasitism/Circular• Not Locally testable• But it IS testable by fragments

Page 40: Crick’s early Hypothesis  Revisited

Interpretation with respect to Crick’s Hypothesis

• Existence of a universal coding frame

• Some families fit the local testability/comma free /BP/circular

• Some families are more susceptible to alternative splicing still they are Testable by Fragments (within the coding sequence)

Page 41: Crick’s early Hypothesis  Revisited

Strict Algorithm

Fw CCw / Fw CCw /

Page 42: Crick’s early Hypothesis  Revisited

Relaxed Algorithm

50SS FF

50SSS FFF

&

Page 43: Crick’s early Hypothesis  Revisited

General Results

• 95.4% success with Strict algorithm

• 94.8% success with Relaxed algorithm

• Distribution of failures (concentrated on some organisms)

• Support the Universal Frame Hypothesis

• Existence of underlying mathematical structures

Page 44: Crick’s early Hypothesis  Revisited

Smallest fragment sizeRelaxed Algorithm

fragment of size 10, window size 2 74% success

fragment of size 60, window size 2590% success

• Keep testable by fragment• Most probable

Page 45: Crick’s early Hypothesis  Revisited

Universal Property

Human - TP53 Gene

ATGGAGGAGCCGCAGTCAGATCCTAGCGTCGAGCCCCCTCTGAGTCAGGAAACATTTTCAGACCTATGGAAACTACTTCCTGAAAACAACGTTCTGTCCCCCTTGCCGTCCCAAGCAATGGATGATTTGATGCTGTCCCCGGACGATATTGAACAATGGTTCACTGAAGACCCAGGTCCAGATGAAGCTCCCAGAATGCCAGAGGCTGCTCCCCGCGTGGCCCCTGCACCAGCAGCTCCTACACCGGCGGCCCCTGCACCAGCCCCCTCCTGGCCCCTGTCATCTTCTGTCCCTTCCCAGAAAACCTACCAGGGCAGCTACGGTTTCCGTCTGGGCTTCTTGCATTCTGGGACAGCCAAGTCTGTGACTTGCACGTACTCCCCTGCCCTCAACAAGATGTTTTGCCAACTGGCCAAGACCTGCCCTGTGCAGCTGTGGGTTGATTCCACACCCCCGCCCGGCACCCGCGTCCGCGCCATGAGA………………………………………………………………………….

Using this gene we are able to find the frame of any other gene.

Ecoli – dgkA Gene……..TCGAATAATACCACTGGATTCACCCGAATTATCAAAGCTTCC…..

Pseudomonas fluorescens – ahcY Gene….TACGGCTGCCGTCACAGCCTGAACGACGCCATCAAGCGCGGC……..

Bos taurus – APOE Gene………..GCTGGGGCCAGCGAGGGTGCCGAGCGCAGCTTGAGCGCCATC…

Sus scrofa - JAK2 Gene……ATTGTAACTATTCATAAGCAAGATGGCAAAAGTCTGGAAAGC……

Pyrococcus – OT3 Gene……CATAGCGTTAACCACTACACCAACAGCGTCGGCAAAATCCTC……

Methanococcus maripaludis – comE Gene….TTTAACAATTACGCACCTATAACTACAGAACAACAACGTGAT……….

Page 46: Crick’s early Hypothesis  Revisited

CONCLUSION

• Provided we extend the notion of Comma-Free to the related notion of Testable By Fragment

Crick’s 1957 Hypothesis is vindicated:

• There exists a universal frame based on a mathematical model

Page 47: Crick’s early Hypothesis  Revisited

Coding vs. Non Coding

Algorithm tells us the most likely coding frame under the assumption that we are in the coding regionNot suitable as such to analyze the non coding region. Need to adapt and refine.

Non coding region contains pseudo genes, gene complements, hypothetical genes, other functional regions in %’ UTR and 3’ UTR…Repeats, and apparently random sequences.Nevertheless we ran an experiment (Augustus) …. 60 pb of transcription vs. translation