Download
promoter prediction in e coli using ann n.
Skip this Video
Loading SlideShow in 5 Seconds..
Promoter Prediction in E.coli using ANN PowerPoint Presentation
Download Presentation
Promoter Prediction in E.coli using ANN

Promoter Prediction in E.coli using ANN

74 Views Download Presentation
Download Presentation

Promoter Prediction in E.coli using ANN

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Promoter Prediction in E.coli using ANN A.Krishnamachari Bioinformatics Centre, JNU chari@mail.jnu.ac.in

  2. Definition of Bioinformatics • Systematic development and application of Computing and Computational solution techniques to biological data to investigate biological process and make novel observations

  3. Research in Biology General approach Bioinformatics era Organism Functions Cell Chromosome DNA Sequences

  4. Genome Sequence intergenic TSS RBS CDS TF -10 -35 TF -> Transcription Factor Sites TSS->Transcription Start Sites RBS -> Ribosome Binding sites CDS - > Coding Sequence (or) Gene

  5. Statement of the problem • Given a set of known sequences pertaining to a specific biological feature , develop a computational method to search for new members or sequences

  6. Computational Methods • Pattern Recognition • Pattern classification • Optimisation Methods

  7. Sequence Analysis AND Prediction Methods • Consensus • Position Weight Matrix (or) Profiles • Machine Learning Methods • Neural Networks • Markov Models • Support Vector Machines • Decision Tree • Optimization Methods Statistical Learning

  8. Promoter- TATA BOX TAA T TA Consensus sequence 49%,54% and 58% 14 sites out of 291 sequences [Lisser and Margalitt] Mismatches but which one?

  9. Relative Entropy Plot - Promoters

  10. Relative Entropy Plot - Random Sequences

  11. Alignment

  12. Describing features using frequency matrices • Goal: Describe a sequence feature (or motif) more quantitatively than possible using consensus sequences • Need to describe how often particular bases are found in particular positions in a sequence feature

  13. Describing features using frequency matrices • Definition: For a feature of length m using an alphabet of ncharacters, a frequency matrixis an n by m matrix in which each element contains the frequency at which a given member of the alphabet is observed at a given position in an aligned set of sequences containing the feature

  14. Frequency matrices (continued) • Three uses of frequency matrices • Describe a sequence feature • Calculate probability of occurrence of feature in a random sequence • Calculate degree of match between a new sequence and a feature

  15. Frequency Matrices, PSSMs, and Profiles • A frequency matrix can be converted to a Position-Specific Scoring Matrix (PSSM) by converting frequencies to scores • PSSMs also called Position Weight Matrixes (PWMs) or Profiles

  16. Methods for converting frequency matrices to PSSMs • Using log ratio of observed to expected where m(j,i) is the frequency of character j observed at position i and f(j) is the overall frequency of character j (usually in some large set of sequences) • Using amino acid substitution matrix (Dayhoff similarity matrix) [see later]

  17. Finding occurrences of a sequence feature using a Profile • As with finding occurrences of a consensus sequence, we consider all positions in the target sequence as candidate matches • For each position, we calculate a score by “looking up” the value corresponding to the base at that position

  18. Alignment

  19. Positions (Columns in alignment) V1 x12 + x21 + x33 + x44 + x52 TAGCT AGTGC if is above a threshold it is a site V1

  20. Building a PSSM Set of Aligned Sequence Features PSSM builder PSSM Expected frequencies of each sequence element

  21. Searching for sequences related to a family with a PSSM Set of Aligned Sequence Features PSSM builder Expected frequencies of each sequence element PSSM Sequences that match above threshold PSSM search Threshold Positions and scores of matches Set of Sequences to search

  22. Consensus sequences vs. frequency matrices • consensus sequence or a frequency matrix which one to use? • If all allowed characters at a given position are equally "good", use IUB codes to create consensus sequence • Example: Restriction enzyme recognition sites • If some allowed characters are "better" than others, use frequency matrix • Example: Promoter sequences

  23. Consensus sequences vs.frequency matrices • Advantages of consensus sequences: smaller description, quicker comparison • Disadvantage: lose quantitative information on preferences at certain locations

  24. Linear Classification Problems Measure2 Measure1

  25. Nonlinear Classification Problem Measure 2 Measure 1

  26. FEATURE EXTRACTION

  27. FEATURE EXTRACTION

  28. (Artificial) Neural Network

  29. What Is A Neural Network ? • A computational construct based on biological neuron • . A neural network can: • Learn by adapting its synaptic weights to changes in the surrounding environments; • handle imprecise, fuzzy, noisy, and probabilistic information; and • generalize from known tasks or examples to unknown ones. • Artificial neural networks (ANNs) attempt to mimic some,or all of these characteristics.

  30. Neural Network • Characterised by: - its pattern of connections between the neurons (Network Architecture) - its method of determining the weights on the connections (training or learning algorithm)

  31. Why Neural Network:Applications -Little or no incomplete understanding of the problem to be solved (very little theory) -Abundant data available

  32. Neural Networks: Applications • Pattern classification • Speech synthesis and recognition • Image compression • Clustering • Medical Diagnosis • Manufacturing

  33. Neural Network:Bioinformatics • Binding sites prediction • Protein Secondary Structure prediction • Protein folds • Micro array data clustering • Gene prediction

  34. Neural Networks • Supervised Learning • Unsupervised Learning

  35. Perceptron Output inputs Layer 2 (output) Layer 1 (input) 1 W1,3 3 W2,3 2 Direction of information flow

  36. Perceptron Summation Operation xi * wij=x1*w1+x2*w2+x3w3….+xnWnj Thresholding functions Output = 0 if x*w < T Output = 1 if x*w >T 1 Output 1 Output 1 Threshold=0 0 0 T

  37. Perceptron Output Input Logistic Transfer function 1 Output = - 1 + e Weight updates W(k+1) = w(k)+ µ[T9k) – w(k)x(k)]x(k) for 0 ≤ k≤ N-1

  38. Learning Concepts • Generally • the target output is 1 for +ve • The target output is 0 for –ve • But practically (0.9, 0.1) combination • Stopping criterion Based on certain epochs or cycles Based on certain error estimates

  39. Perceptron Nucleic Acid A T G C 1 2 Position In a sequence Of K nucleotides K-1 K

  40. Bit-Coding • let the following binary values represent each base • A="0001 • C="0010 • G="0100 • T="1000 • then • G = 4 • A or C = "0011 = 3 • A,G or T = "1101 = 13 • etc.

  41. NETWORK Nucleic Acid A T G C 1 A 0 0 0 1 Wi,j 2 G 0 1 0 0 Position In a sequence Of K nucleotides K-1 G 0 1 0 0 K T 1 0 0 0

  42. Learning Model Test set Positive Model Negative Model P=50 N=50 TP + TN =100 Note: Training and Test sequences are fixed length

  43. Learning Model Training Set Test Set 1 2 …………50 1 2 …………50 n=10 N=500

  44. Learning Model TRAINING C G T A G C T A T A G T G G G T T T A A A C C C A A G A A T T A T G G A A T T T G G A A G T T T A G G A T A G C A C A G G A T A A G G C C T A G A T A T T T A T G C A T G A G A T G Prediction Method Output TEST C C T G A A C T G A G A T G A T A T A T A A G T G A A A T T C C G Input

  45. Multilayer Perceptron Input Layer -1 Hidden layer Output Layer