Skip to main content

Table 4 Examples of results of the First Trial

From: A fuzzy method for RNA-Seq differential expression analysis in presence of multireads

GENE: OTTHUMG00000031027|HLA-DRB5

FUZZY COUNTS

Control

Case

Under-expr.

Same-expr.

Over-expr.

Tr[18,24,26,428]

Tr[3732,4551,4666,6386]

1

0.003

0

ESTIMATED COUNTS

Control

Case

STATS

p-value

Log2(FC)

 TopHat-unique (RPKM)

0.2

250.7

cuffdiff

0.9996

10.3

 TopHat-rescue (RPKM)

0.2

219.5

cuffdiff

0.9985

10.5

 Uniquely mapping

18

3732

FDR

0

7.6

 RSEM

23

4665

FDR

0

7.6

 Centroid

25

4607

FDR

0

8.1

GENE: OTTHUMG00000036468|TTTY15

FUZZY COUNTS

Control

Case

Under-expr.

Same-expr.

Over-expr.

Tr[0,1,1,17]

Tr[43,43,43,52]

0.024

0.986

0.007

ESTIMATED COUNTS

Control

Case

STATS

p-value

Log2(FC)

 TopHat-unique (RPKM)

0

0.4

cuffdiff

0.0218

Inf

 TopHat-rescue (RPKM)

0

0.4

cuffdiff

0.029

Inf

 Uniquely mapping

0

43

FDR

1E-14

5.5

 RSEM

1

43

FDR

2E-14

4.5

 Centroid

1

43

FDR

2E-14

4.9

GENE: OTTHUMG00000129909|IGJ

FUZZY COUNTS

Control

Case

Under-expr.

Same-expr.

Over-expr.

Tr[0,0,291,318]

Tr[0,0,1227,1246]

1

1

0.217

ESTIMATED COUNTS

Control

Case

STATS

p-value

Log2(FC)

 TopHat-unique (RPKM)

16.4

84.9

cuffdiff

0.0218

2.4

 TopHat-rescue (RPKM)

17.9

88.4

cuffdiff

0.029

2.3

 Uniquely mapping

0

0

FDR

1

0.0

 RSEM

250

1165

FDR

2E-208

2.2

 Centroid

146

614

FDR

9E-103

2.7

  1. The first gene, HLA-DRB5, despite the abundant presence of multireads in its read counts, shows a certain result, with an maximum under-expression possibility and very low same-expression possibility. This results is not confirmed by cuffdiff, evenif it computes a high FC. The second gene, TTTY15 result is not reliable because it has low read counts, but it is considered as a DE result by all the tools. Our method higlights the result as false positive, with a high same-expression value, because of the low mean expression values obtained. The third example, the IGJ gene, is a possible false positive. Its counts are quite low and its fuzzy read counts are mostly overlapping. The pvalues obtained by the other tools are confirmed by a high under-expression possibility, but the risk of having a false positive is pointed out by a high same-expression possibility
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us