| .names file created by George John, October 1994
| Processing:
| * A,C,T,G -> 001,010,100,000 Seems biased against systems that can handle
| categorical attributes
|
|
|
|1. TITLE:
| DNA Dataset (STATLOG version) - Primate splice-junction gene sequences (DNA)
| with associated imperfect domain theory
|
| PROBLEM DESCRIPTION
| Splice junctions are points on a DNA sequence at which `superfluous' DNA is
| removed during the process of protein creation in higher organisms. The
| problem posed in this dataset is to recognize, given a sequence of DNA, the
| boundaries between exons (the parts of the DNA sequence retained after
| splicing) and introns (the parts of the DNA sequence that are spliced
| out).
|
| PURPOSE
| This problem consists of two subtasks: recognizing exon/intron
| boundaries (referred to as EI sites), and recognizing intron/exon boundaries
| (IE sites). (In the biological community, IE borders are referred to
| a ``acceptors'' while EI borders are referred to as ``donors''.)
|
|2. USE IN STATLOG
|
| 2.1- Testing Mode
| Train & Test
|
| 2.2- Special Preprocessing
| Yes
|
| 2.3- Test Results
|
| Algorithm Sucess Rate
| ========= ===========
| Radial 95.90
| Dipol92 95.200
| Alloc80 94.300
| QuaDisc 94.100
| Discrim 94.100
| LogDisc 93.900
| Bayes 93.200
| Castle 92.800
| IndCart 92.700
| C4.5 92.400
| Cart 91.500
| BackProp 91.200
| BayTree 90.500
| Cn2 90.500
| Ac2 90.000
| NewId 90.000
| Cal5 86.900
| Itrule 86.500
| Smart 85.900
| KNN 84.500
| Kohonen 66.10
| Default 52.000
| LVQ 0.000
| Cascade 0.000
|
|3. SOURCES and PAST USAGE
| 3.1 SOURCES
| (a) Creators:
| - all examples taken from Genbank 64.1 (ftp site: genbank.bio.net)
| - categories "ei" and "ie" include every "split-gene"
| for primates in Genbank 64.1
| - non-splice examples taken from sequences known not to include
| a splicing site
| (b) Donor: G. Towell, M. Noordewier, and J. Shavlik,
| {towell,shavlik}@cs.wisc.edu, noordewi@cs.rutgers.edu
| (c) Date received: 1/1/92
|
| The StaLog dna dataset is a processed vesrion of the Irvine
| database described below. The main difference is that the
| symbolic variables representing the nucleotides (only A,G,T,C)
| were replaced by 3 binary indicator variables. Thus the original
| 60 symbolic attributes were changed into 180 binary attributes.
| The names of the examples were removed. The examples with
| ambiguities were removed (there was very few of them, 4).
| The StatLog version of this dataset was produced by Ross King
| at Strathclyde University. For original details see the Irvine
| database documantation.
|
| The nucleotides A,C,G,T were given indicator values as follows
|
| A -> 1 0 0
| C -> 0 1 0
| G -> 0 0 1
| T -> 0 0 0
|
| The class values are
| ei -> 1
| ie -> 2
| n -> 3
| 3.2 PAST USAGE
|
| (a) machine learning:
| -- M. O. Noordewier and G. G. Towell and J. W. Shavlik, 1991;
| "Training Knowledge-Based Neural Networks to Recognize Genes in
| DNA Sequences". Advances in Neural Information Processing Systems,
| volume 3, Morgan Kaufmann.
|
| -- G. G. Towell and J. W. Shavlik and M. W. Craven, 1991;
| "Constructive Induction in Knowledge-Based Neural Networks",
| In Proceedings of the Eighth International Machine Learning
| Workshop, Morgan Kaufmann.
|
| -- G. G. Towell, 1991;
| "Symbolic Knowledge and Neural Networks: Insertion, Refinement, and
| Extraction", PhD Thesis, University of Wisconsin - Madison.
|
| -- G. G. Towell and J. W. Shavlik, 1992;
| "Interpretation of Artificial Neural Networks: Mapping
| Knowledge-based Neural Networks into Rules", In Advances in Neural
| Information Processing Systems, volume 4, Morgan Kaufmann.
|
| (b) attributes predicted: given a position in the middle of a window
| 60 DNA sequence elements (called "nucleotides" or "base-pairs"),
| decide if this is a
| a) "intron -> exon" boundary (ie) [These are sometimes called "donors"]
| b) "exon -> intron" boundary (ei) [These are sometimes called "acceptors"]
| c) neither (n)
| (c) Results of study indicated that machine learning techniques (neural
| networks, nearest neighbor, contributors' KBANN system) performed as
| well/better than classification based on canonical pattern matching
| (method used in biological literature).
|
| HISTORY
| This dataset has been developed to help evaluate a "hybrid" learning
| algorithm (KBANN) that uses examples to inductively refine preexisting
| knowledge. Using a "ten-fold cross-validation" methodology on 1000
| examples randomly selected from the complete set of 3190, the following
| error rates were produced by various ML algorithms (all experiments
| run at the Univ of Wisconsin, sometimes with local implementations
| of published algorithms).
|
| System Neither EI IE
| ---------- ------- ----- -----
| KBANN 4.62 7.56 8.47
| BACKPROP 5.29 5.74 10.75
| PEBLS 6.86 8.18 7.55
| PERCEPTRON 3.99 16.32 17.41
| ID3 8.84 10.58 13.99
| COBWEB 11.80 15.04 9.46
| Near. Neighbor 31.11 11.65 9.09
|
| Type of domain: non-numeric, nominal (one of A, G, T, C)
|
|*************************************************************
|
|4. DATASET DISCRIPTION
| NUMBER OF EXAMPLES:
| 3186
|
| Train 2000
| Test 1186
|
| NUMBER OF CLASSES:
| 3 (one of 1,2,3)
|
| Distribution of classes
| Class Train Test
| ------------------------------------
| 1 464 (23.20%) 303 (25.55%)
| 2 485 (24.25%) 280 (23.61%)
| 3 1051 (52.55%) 603 (50.84%)
|
| NUMBER OF ATTRIBUTES:
| 180 binary indicator variables
|
| Hint. Much better performance is generally observed if attributes
| closest to the junction are used.
| In the StatLog version, this means using
| attributes A61 to A120 only.
|
|
|CONTACTS
| statlog-adm@ncc.up.pt
| bob@stams.strathclyde.ac.uk
|
|
|================================================================================
|;little lisp function to generate names:
|(defun atts ()
| (let ((i 1))
| (while (<= i 180)
| (insert (format "A%s: continuous.\n" i))
| (setq i (+ 1 i)))))
1,2,3. | classes
A0: 0,1.
A1: 0,1.
A2: 0,1.
A3: 0,1.
A4: 0,1.
A5: 0,1.
A6: 0,1.
A7: 0,1.
A8: 0,1.
A9: 0,1.
A10: 0,1.
A11: 0,1.
A12: 0,1.
A13: 0,1.
A14: 0,1.
A15: 0,1.
A16: 0,1.
A17: 0,1.
A18: 0,1.
A19: 0,1.
A20: 0,1.
A21: 0,1.
A22: 0,1.
A23: 0,1.
A24: 0,1.
A25: 0,1.
A26: 0,1.
A27: 0,1.
A28: 0,1.
A29: 0,1.
A30: 0,1.
A31: 0,1.
A32: 0,1.
A33: 0,1.
A34: 0,1.
A35: 0,1.
A36: 0,1.
A37: 0,1.
A38: 0,1.
A39: 0,1.
A40: 0,1.
A41: 0,1.
A42: 0,1.
A43: 0,1.
A44: 0,1.
A45: 0,1.
A46: 0,1.
A47: 0,1.
A48: 0,1.
A49: 0,1.
A50: 0,1.
A51: 0,1.
A52: 0,1.
A53: 0,1.
A54: 0,1.
A55: 0,1.
A56: 0,1.
A57: 0,1.
A58: 0,1.
A59: 0,1.
A60: 0,1.
A61: 0,1.
A62: 0,1.
A63: 0,1.
A64: 0,1.
A65: 0,1.
A66: 0,1.
A67: 0,1.
A68: 0,1.
A69: 0,1.
A70: 0,1.
A71: 0,1.
A72: 0,1.
A73: 0,1.
A74: 0,1.
A75: 0,1.
A76: 0,1.
A77: 0,1.
A78: 0,1.
A79: 0,1.
A80: 0,1.
A81: 0,1.
A82: 0,1.
A83: 0,1.
A84: 0,1.
A85: 0,1.
A86: 0,1.
A87: 0,1.
A88: 0,1.
A89: 0,1.
A90: 0,1.
A91: 0,1.
A92: 0,1.
A93: 0,1.
A94: 0,1.
A95: 0,1.
A96: 0,1.
A97: 0,1.
A98: 0,1.
A99: 0,1.
A100: 0,1.
A101: 0,1.
A102: 0,1.
A103: 0,1.
A104: 0,1.
A105: 0,1.
A106: 0,1.
A107: 0,1.
A108: 0,1.
A109: 0,1.
A110: 0,1.
A111: 0,1.
A112: 0,1.
A113: 0,1.
A114: 0,1.
A115: 0,1.
A116: 0,1.
A117: 0,1.
A118: 0,1.
A119: 0,1.
A120: 0,1.
A121: 0,1.
A122: 0,1.
A123: 0,1.
A124: 0,1.
A125: 0,1.
A126: 0,1.
A127: 0,1.
A128: 0,1.
A129: 0,1.
A130: 0,1.
A131: 0,1.
A132: 0,1.
A133: 0,1.
A134: 0,1.
A135: 0,1.
A136: 0,1.
A137: 0,1.
A138: 0,1.
A139: 0,1.
A140: 0,1.
A141: 0,1.
A142: 0,1.
A143: 0,1.
A144: 0,1.
A145: 0,1.
A146: 0,1.
A147: 0,1.
A148: 0,1.
A149: 0,1.
A150: 0,1.
A151: 0,1.
A152: 0,1.
A153: 0,1.
A154: 0,1.
A155: 0,1.
A156: 0,1.
A157: 0,1.
A158: 0,1.
A159: 0,1.
A160: 0,1.
A161: 0,1.
A162: 0,1.
A163: 0,1.
A164: 0,1.
A165: 0,1.
A166: 0,1.
A167: 0,1.
A168: 0,1.
A169: 0,1.
A170: 0,1.
A171: 0,1.
A172: 0,1.
A173: 0,1.
A174: 0,1.
A175: 0,1.
A176: 0,1.
A177: 0,1.
A178: 0,1.
A179: 0,1.