Display Settings:

Items per page

Results: 11

1.
Fig. 5.

Fig. 5. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

A false positive BLAT alignment of a 120 base read of mouse retina. BLAT has excessively fragmented the read and aligned it to a low complexity region.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
2.
Fig. 9.

Fig. 9. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

The sensitivity and positive predictive value (PPV) at the individual read level. MapSplice and RUM have the highest overall sensitivity, while all algorithms have PPV ~ 65%.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
3.
Fig. 1.

Fig. 1. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

BEERS simulator workflow. Genes are chosen at random from a master pool, polymorphisms and novel splice forms are introduced, and then reads are generated in a six step cycle, as shown.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
4.
Fig. 3.

Fig. 3. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

The RUM workflow. Reads are first mapped with Bowtie against the genome and transcriptome. This information is merged and non-mappers are sent to BLAT. BLAT and Bowtie mappings are merged for the final alignments. Features are quantified and coverage and junction files are produced.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
5.
Fig. 2.

Fig. 2. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

Gapped alignment using BLAT. BLAT alignments (segments in black) of a mouse retina 108 base read that spans three exon/exon junctions. The second junction is unannotated, according to the USCS annotation track shown in blue.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
6.
Fig. 10.

Fig. 10. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

Processor time required for analysis of simulated datasets. The processor time required for each of the algorithms tested to analyze the first (A) and second (B) simulated datasets is shown. Data are mean ± SEM. The values from which these graphs are derived are shown in Supplementary Table 2. Algorithms were run on 64 bit Linux Debian with 2.6 GHz processors.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
7.
Fig. 7.

Fig. 7. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

Representative coverage plots demonstrating the effect of a two base deletion on alignments with the algorithms indicated. Reads were aligned using RUM, the individual BLAT and Bowtie components of RUM, and 10 currently available alignment algorithms. The TRUTH coverage plot (top) represents the true alignment of the reads containing the two-base deletion (arrow). RUM and several other algorithms were able to correctly align these reads. Note that TopHat, SpliceMap, Bowtie and Soap, which do not identify indels, fail to accurately align reads to these regions.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
8.
Fig. 8.

Fig. 8. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

Comparison of accuracies near junctions on BEERS-generated data. The true junctions are shown in black at the bottom of the figure. Reads mapping to the region of the simulated annotation track (bottom) were aligned using RUM, the individual components of RUM and the 10 currently available alignment algorithms indicated. The TRUTH coverage plot represents the true alignment of the simulated reads. There are five characteristic splice junction sites (1–5) that indicate varying accuracy of the alignment algorithms. BLAT- and the BLAT-based algorithms RUM and RUM-Genome provide the most accurate resolution of the depicted junctions. GSNAP detects the five junctions, and also displays inaccurate alignment of reads in the intron near junction #2.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
9.
Fig. 4.

Fig. 4. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

This illustrates a hypothetical case where it is difficult to resolve the transcriptome and genome mappings. (A) Shows how a read aligns to the genome. It spans an exon and erroneously extends one or two bases on either side into the intron. (B) Shows how the same read maps to the transcriptome. In this case, the few terminal bases map to the adjacent exons. (C) Shown in red is the alignment of the paired-end read, which has aligned to the intron. Even if all bases of all alignments are identities, if all we had was the information in (A) and (B) we would likely preference the transcriptome alignment (B). If we have the information in (C) then we would preference the genome mapping in (A) on the right, but it becomes a difficult judgment on the left, given that the retention of selective introns and partial introns is frequently observed.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
10.
Fig. 6.

Fig. 6. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

Accuracy statistics for analyses of simulated datasets. (A and B) Simulated dataset 1. (C and D) Simulated dataset 2. Test 1 has low polymorphism and error rates, while Test 2 has moderate polymorphism and error rates. In (A) and (C), the bars show the base-wise accuracy (the percent of bases that aligned and to the right location). (B) and (D) Show the accuracy of the junction calls, dark bars show the false positive (FP) rate and light bars show the false negative (FN) rate. The algorithms are sorted in (A) and (C) by accuracy and in (B) and (D) by the sum of the FP and FN rates. Results are mean ± SEM over the three replicate simulated datasets for each test. There is a considerable drop-off in accuracy seen in Test 2 for the algorithms that do not align across indels (SpliceMap, TopHat and Bowtie). The base-wise accuracy and the FP and FN rates on junction calls are taken in conjunction to determine the overall effectiveness of an algorithm. Based on these results, we conclude that GSNAP, MapSplice and RUM are the ones that are most viable for RNA-Seq alignment.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.
11.
Fig. 11.

Fig. 11. From: Comparative analysis of RNA-Seq alignment algorithms and the RNA-Seq unified mapper (RUM).

Validation of novel splice junctions detected by RUM. Exon junctions detected by RUM are displayed as a track using the UCSC Genome Browser. The reads with annotated junctions are displayed in blue; reads with novel junctions are shown in green. The depth of uniquely mapped sequence reads is shown in the Coverage Plot in red. The BLAT aligned Sanger sequenced reads from RT–PCR products are shown in black under the coverage plot. Annotated Ensemble and UCSC genes are indicated at the bottom of the images. (A) RUM aligned five RNA-seq reads cleanly across a putative novel junction between exons 29 and 31 of the Usp32 gene, compared with 525 and 349 reads that detected the 5 and 3 ends of annotated exon 30, respectively. RT–PCR and Sanger sequencing in independent biological samples confirmed the presence of the mRNA lacking exon 30. (B) The 47 reads aligned to a putative novel alternate splice junction at the 5 end of 7th and final exon of Bcl9, while 144 reads aligned to the known junction. The novel junction removes 36 bases, and 12 amino acids in frame from the coding sequence. RT–PCR and Sanger sequencing in independent biological samples confirmed the presence of the mRNA with this novel junction. (C) An abundant putative novel exon was detected between exons 50 and 51 of Mll2 gene. In addition to detection by the RUM junction track, this exon is also evident in the coverage plot. The 48-bp novel exon is predicted to add 16 amino acids in frame to the Mll2 protein. RT–PCR and Sanger sequencing in independent biological samples confirmed the presence of this novel transcript. (D) A low abundance putative novel exon was detected between exons 2 and 3 of the Gtf2a1 gene. RT–PCR and Sanger sequencing in independent biological samples validated the expression of the novel Gtf2a1 transcript containing this 81 bp exon.

Gregory R. Grant, et al. Bioinformatics. 2011 September 15;27(18):2518-2528.

Display Settings:

Items per page

Supplemental Content

Recent activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...
Write to the Help Desk