IRR prediction

- Prediction of ionizing radiation resistance in bacteria using a multiple-instance learning model
- (Accepted in Journal of Computational Biology)
Authors | Datasets | Results | Download

Authors:

- Sabeur Aridhi
  
  Tel: +39 04 61 28 39 92
  Email: sabeur.aridhi@isima.fr
  Web page: http://fc.isima.fr/~aridhi/
  DISI, Department of Information Engineering and Computer Science
  University of Trento
  Via Sommarive 9, 38123 Trento, Italy
- Haïtham Sghaier
  
  Email: sghaier.haitham@gmail.com
  Google Scholar
  
  National Center for Nuclear Sciences and Technology (CNSTN)
  Sidi Thabet Technopark
  Ariana 2020, Tunisia
- Manel Zoghlami
  
  Tel: +33 4 73 40 76 29
  Email: manel.zoghlami@gmail.com
  
  LIMOS, UMR 6158
  Blaise Pascal University - Clermont University
  BP 10125, 63173, Clermont Ferrand, France
- Mondher Maddouri
  
  Email: maddourimondher@yahoo.fr
  
  Faculty of sciences of Tunis, LIPAH
  University of Tunis El Manar
  1060, Tunis, Tunisia
- Engelbert Mephu Nguifo
  
  Tel: +33 4 73 40 76 29
  Email: mephu@isima.fr
  Web page: http://www.isima.fr/~mephu/
  
  LIMOS, UMR 6158
  Blaise Pascal University - Clermont University
  BP 10125, 63173, Clermont Ferrand, France

Datasets:

General description

For our experiments, we constructed a database containing 14 IRRB and 14 IRSB. Table 1 presents the used Bacteria.
Each bacterium contains 25 to 31 proteins implicated in basal DNA repair in IRRB (see Table 2).

Data sources

Proteins of the bacterium Deinococcus radiodurans were downloaded from the UniProt web site. http://www.uniprot.org/uniprot/
prfectBLAST tool was used to identify orthologous proteins of the other bacteria. Download prfectBLAST
Proteomes of other bacteria were downloaded from the NCBI FTP web site. http://www.ncbi.nlm.nih.gov/Ftp/

Used bacteria and proteins.

Table. 1: IRRB and IRSB learning set.

Phenotype	ID	Bacterium	Phylogenetic group	D10
IRRB	B1	Chroococcidiopsis thermalis PCC 7203	Cyanobacteria	4*
	B2	Deinococcus deserti VCD115	Deinococcus-Thermus	>7.5
	B3	Deinococcus geothermalis DSM 11300	Deinococcus-Thermus	10-16
	B4	Deinococcus gobiensis I 0	Deinococcus-Thermus	12.7
	B5	Deinococcus maricopensis DSM 21211	Deinococcus-Thermus	~11
	B6	Deinococcus proteolyticus MRP	Deinococcus-Thermus	>15
	B7	Deinococcus radiodurans R1	Deinococcus-Thermus	10
	B8	Geodermatophilus obscurus DSM 43160	Actinobacteria	9
	B9	Kineococcus radiotolerans SRS30216	Actinobacteria	2
	B10	Kocuria rhizophila DC2201	Actinobacteria	2**
	B11	Methylobacterium radiotolerans JCM 2831	Proteobacteria	1
	B12	Modestobacter marinus	Actinobacteria	6
	B13	Rubrobacter xylanophilus DSM 9941	Actinobacteria	5.5
	B14	Truepera radiovictrix DSM 17093	Deinococcus-Thermus	>5
IRSB	B15	Brucella abortus S19	Proteobacteria	0.34
	B16	Escherichia coli B REL606	Proteobacteria	0.7
	B17	Escherichia coli str. K-12 substr. DH10B	Proteobacteria	0.7
	B18	Neisseria gonorrhoeae FA 1090	Proteobacteria	0.07-0.125
	B19	Neisseria gonorrhoeae TCDC NG08107	Proteobacteria	0.07-0.125
	B20	Pseudomonas putida S16	Proteobacteria	0.25
	B21	Shewanella oneidensis MR-1	Proteobacteria	0.07
	B22	Shigella dysenteriae1617	Proteobacteria	0.22
	B23	Thermus thermophilus HB27	Deinococcus-Thermus	0.8
	B24	Thermus thermophilus HB8	Deinococcus-Thermus	0.8***
	B25	Thermus thermophilus JL-18	Deinococcus-Thermus	0.8***
	B26	Thermus thermophilus SG0.5JP17-16	Deinococcus-Thermus	0.8***
	B27	Vibrio parahaemolyticus RIMD 2210633	Proteobacteria	0.03-0.06
	B28	Yersinia enterocolitica 8081	Proteobacteria	0.1-0.21

*for Chroococcidiopsis spp
**for Kocuria rosea
***for T. thermophilus HB27

Table. 2: Replication, repair and recombination proteins.

ID	Protein	Function
P1	Hypothetical DNA polymerase	DNA polymerase
P2	DNA polymerase III, α subunit
P3	DNA-directed DNA polymerase
P4	DNA polymerase III, τ/γ subunit
P5	Single-stranded DNA-binding protein	Replication complex
P6	Replicative DNA helicase
P7	DNA primase
P8	DNA gyrase, subunit B
P9	DNA topoisomerase I
P10	DNA gyrase subunit A
P11	smf protein	Other DNA-associated proteins
P12	Endonuclease III
P13	Holliday junction resolvase
P14	Formamidopyrimidine-DNA glycosylase
P15	Holliday junction DNA helicase
P16	RecF protein
P17	DNA repair protein radA
P18	Holliday junction binding protein
P19	Excinuclease ABC, subunit C
P20	DNA repair protein RecN
P21	Transcription-repair coupling factor
P22	Excinuclease ABC, subunit A
P23	DNA helicase II
P24	DNA helicase RecG
P25	Exonuclease SbcD, putative
P26	Exonuclease SbcC
P27	Ribonuclease HII
P28	Excinuclease ABC, subunit B
P29	A/G-specific adenine glycosylase
P30	RecA protein
P31	DNA-3-methyladenine glycosidase II, putative

Results:

The computations were carried out on a i7 CPU 2.49 GHz PC with 6 GB of memory, operating on Linux Ubuntu. In the classification process, we used the Leave-One-Out (LOO) evaluation technique.

Table 3. Experimental results of MIL-ALIGN with LOO-based evaluation technique.

Used proteins	Aggregation method	Accuracy (%)	Sensitivity (%)	Specificity (%)
All proteins	SMS	92.80	92.80	92.80
All proteins	WAMS	89.2	92.30	86.6
DNA Polymerase proteins	SMS	89.2	92.30	86.6
DNA Polymerase proteins	WAMS	89.2	92.30	86.6
Replication complex proteins	SMS	92.80	92.80	92.80
Replication complex proteins	WAMS	92.80	92.80	92.80
Other DNA-associated proteins	SMS	92.80	92.80	92.80
Other DNA-associated proteins	WAMS	92.80	92.80	92.80

Table. 4: Percentage of successful predictions using MIL.

Phenotype	Bacterium ID	Successful predictions (%)
IRRB	B1	100
	B2	100
	B3	100
	B4	100
	B5	100
	B6	100
	B7	100
	B8	100
	B9	100
	B10	100
	B11	0
	B12	100
	B13	100
	B14	62.5*
IRSB	B15	0
	B16	100
	B17	100
	B18	100
	B19	100
	B20	100
	B21	100
	B22	100
	B23	100
	B24	100
	B25	100
	B26	100
	B27	100
	B28	100

*successfully classified bacterium using: (1) all proteins with SMS aggregation method (2) replication complex proteins with SMS and WAMS aggregation methods and (3) other DNAassociated proteins with SMS and WAMS aggregation methods.

Table. 5: Learning results with the traditional setting of machine learning.

Protein ID	Accuracy (%)	Sensitivity (%)	Specificity (%)
P1	85.7	100	77.7
P2	89.2	92.3	86.6
P3	82.1	90.9	76.4
P4	89.2	92.3	86.6
P5	89.2	92.3	86.6
P6	89.2	92.3	86.6
P7	89.2	92.3	86.6
P8	78.5	83.3	75
P9	89.2	92.3	86.6
P10	89.2	92.3	86.6
P11	89.2	92.3	86.6
P12	89.2	92.3	86.6
P13	78.5	90	72.2
P14	89.2	92.3	86.6
P15	85.7	91.6	81.2
P16	89.2	92.3	86.6
P17	85.7	91.6	81.2
P18	85.7	91.6	81.2
P19	89.2	92.3	86.6
P20	85.7	91.6	81.2
P21	85.7	91.6	81.2
P22	89.2	92.3	86.6
P23	89.2	92.3	86.6
P24	89.2	92.3	86.6
P25	85.7	91.6	81.2
P26	82.1	90.9	76.4
P27	82.1	100	73.6
P28	89.2	92.3	86.6
P29	78.5	90	72.2
P30	89.2	92.3	86.6
P31	78.5	78.5	78.5

Download:

MIL-ALIGN is used to predict bacterial ionizing radiation resistance using a multiple-instance learning model. It runs on a Windows or a UNIX platform that contains a Java Runtime Environment (JRE).
MIL-ALIGN (64-bit) is downloadable here.
MIL-ALIGN (32-bit) is downloadable here.
You can download the dataset used in the experiments of the paper here.
Instructions describing how to use MIL-ALIGN can be found in the ReadMe file. You can get a copy from here.

Number of visitors : 3001