Center for Machine Learning and Intelligent Systems
About  Citation Policy  Donate a Data Set  Contact

Repository Web            Google
View ALL Data Sets

× Check out the beta version of the new UCI Machine Learning Repository we are currently testing! Contact us if you have any issues, questions, or concerns. Click here to try out the new site.

M. Tuberculosis Genes Data Set
Download: Data Folder, Data Set Description

Abstract: Data giving characteristics of each ORF (potential gene) in the M. tuberculosis bacterium. Sequence, homology (similarity to other genes) and structural information, and function (if known) are provided

Data Set Characteristics:  


Number of Instances:




Attribute Characteristics:


Number of Attributes:


Date Donated


Associated Tasks:


Missing Values?


Number of Web Hits:



Ross D. King
Department of Computer Science,
University of Wales Aberystwyth,
SY23 3DB, Wales
rdk '@'

Data Set Information:

The data was collected from several sources, including the Sanger Centre ([Web Link]) and SWISSPROT ([Web Link]). Structure prediction was made by PROF ([Web Link]). Homology search was made by PSI-BLAST ([Web Link]).

The data is in Datalog format. Missing values are not explicit, but some genes have more relationships than others.


M. tuberculosis genes (ORFs) are related to each other by the predicate tb_to_tb_evalue(TBNumber,E-value). They are related to other (SWISSPROT) proteins by the predicate e_val(AccNo,E-value). All the data for a single gene (ORF) is enclosed between delimiters of the form:


Other Relevant Information:

The gene functional classes are in a hierarchy. See [Web Link].

There are two datalog files: and


Lists classes and ORF functions. Lines are of the following form:

class([1,0,0,0],"Small-molecule metabolism ").
class([1,1,0,0],"Degradation ").
class([1,1,1,0],"Carbon compounds ").

Arguments are a list of 4 numbers (describing class at the 4 different levels), followed by a string class description. For example,


Arguments are ORF number, list of 4 class numbers, gene name (or null if no gene name) in single quotes, ORF description in double quotes.


Data for each ORF (gene) is delimited by


where X is the ORF number. Other predicates are as follows (examples):

tb_protein(X). % X is gene number
function(2,1,5,0,'gyrA','DNA gyrase subunit A'). % 4 levels of functional hierarchy, gene name, description
coding_region(7302,9815). % start,end. integers
tb_mol_wt(19934). % integer
access(1,e,20). % int (position), {e,i,b}, int (length)
access_exposed(1,20). % int (position), int (length)
access_intermediate(26,1). % int (position), int (length)
access_burried(1,2). % int (position), int (length)
access_dist(b,42.8). % {e,i,b}, float (percentage)
sec_struc(1,c,23). % int (position), {a,b,c}, int (length)
sec_struc_coil(1,23). % int (position), int (length)
sec_struc_alpha(1,15). % int (position), int (length)
sec_struc_beta(1,6). % int (position), int (length)
struc_dist(a,32.1). % {a,b,c}, float (percentage)
sec_struc_conf(78.8). % float (confidence)
sec_struc_conf_alpha(88.9). % float (confidence)
sec_struc_conf_beta(58.0). % float (confidence)
sec_struc_conf_coil(77.7). % float (confidence)
psi_sequences_found(1,7). % how many found, which iteration
psi_sequences_found_again(2,7). % how many found, which iteration
psi_sequences_found_new(2,0). % how many found, which iteration
amino_acid_ratio(a,11.2). % amino acid letter, float
amino_acid_pair_ratio(a,c,0.0). % amino acid letter, amino acid letter, float (out of 1000, ie 2.8 = 0.28%)
sequence_length(187). % integer
tb_to_tb_evalue(tb3671,1.100000e-01). % ORF number, e-value (double)
e_val(p35925,7.0e-59). % SWISSPROT accession no, e-value (double)
species(p35925,'streptomyces_coelicolor'). % SWISSPROT acc no, string
classification(p35925,bacteria). % SWISSPROT acc no, name
mol_wt(p35925,19772). % SWISSPROT acc no, integer
keyword(p35925,'hypothetical_protein'). % SWISSPROT acc no, string
db_ref(p35925,embl,l27063,g436026,null). % SWISSPROT acc no, db id, primary id, secondary id, status id
signalip(c,35,no). % {c,y,s}, int (signal peptide c/y/s score), yes/no
signalip(ss,1,34,no). % ss, int, int, yes/no
signalip(cleavage,59,60). % cleavage, int/null, int/null
hydro_cons(-0.498,-0.474,0.624,3.248,0.278). % double, double, double, double, double
gene_name(p41514,'gyrb'). % SWISSPROT acc no, string

Attribute Information:


Relevant Papers:

King, R. and Karwath, A. and Clare, A. and Dehaspe, L. (2000). Genome Scale Prediction of Protein Functional Class from Sequence Using Data Mining, In Proceedings of the Seventh ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.
[Web Link]

King, R. and Karwath, A. and Clare, A. and Dehaspe, L. (2000). Accurate prediction of protein functional class in the M. tuberculosis and E. coli genomes using data mining, Comparative and Functional Genomics, 17, pp 283--293.
[Web Link]

Citation Request:

Usage Restrictions:

Copyright 2000 by R. D. King, A. Karwath, A. Clare, L. Dehaspe

There are no restrictions. This data is provided "as is" and without any express or implied warranties, including, without limitation, the implied warranties of merchantibility and fitness for a particular purpose.

Citation Requests:

Please cite King~et. al (2000).


This work was supported by the following grants: G78/6609, BIF08765, GR/L62849 and by PharmaDM, Ambachtenlaan, 54/D, B-3001 Leuven, Belgium.

Supported By:

 In Collaboration With:

About  ||  Citation Policy  ||  Donation Policy  ||  Contact  ||  CML