Limits...
Comparison of Analysis Tools for miRNA High Throughput Sequencing Using Nerve Crush as a Model.

Metpally RP, Nasser S, Malenica I, Courtright A, Carlson E, Ghaffari L, Villa S, Tembe W, Van Keuren-Jensen K - Front Genet (2013)

Bottom Line: We also found that samples with very low starting amounts of RNA (mouse plasma) made high depth of mature miRNA coverage more difficult to obtain.We explored the changes in total mapped reads and differential expression results generated by three different software packages: miRDeep2, miRNAKey, and miRExpress and two different analysis packages, DESeq and EdgeR.We also examine the accuracy of using miRDeep2 to predict novel miRNAs and subsequently detect them in the samples using qRT-PCR.

View Article: PubMed Central - PubMed

Affiliation: Collaborative Bioinformatics Center, Translational Genomics Research Institute Phoenix, AZ, USA.

ABSTRACT
Recent advances in sample preparation and analysis for next generation sequencing have made it possible to profile and discover new miRNAs in a high throughput manner. In the case of neurological disease and injury, these types of experiments have been more limited. Possibly because tissues such as the brain and spinal cord are inaccessible for direct sampling in living patients, and indirect sampling of blood and cerebrospinal fluid are affected by low amounts of RNA. We used a mouse model to examine changes in miRNA expression in response to acute nerve crush. We assayed miRNA from both muscle tissue and blood plasma. We examined how the depth of coverage (the number of mapped reads) changed the number of detectable miRNAs in each sample type. We also found that samples with very low starting amounts of RNA (mouse plasma) made high depth of mature miRNA coverage more difficult to obtain. Each tissue must be assessed independently for the depth of coverage required to adequately power detection of differential expression, weighed against the cost of sequencing that sample to the adequate depth. We explored the changes in total mapped reads and differential expression results generated by three different software packages: miRDeep2, miRNAKey, and miRExpress and two different analysis packages, DESeq and EdgeR. We also examine the accuracy of using miRDeep2 to predict novel miRNAs and subsequently detect them in the samples using qRT-PCR.

No MeSH data available.


Related in: MedlinePlus

The correlation of a subset of reads, beginning with 100,000 with the full number of mapped reads, 3.5 million.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3585423&req=5

Figure 6: The correlation of a subset of reads, beginning with 100,000 with the full number of mapped reads, 3.5 million.

Mentions: In addition to examining how the incremental increase in coverage altered the detection of new miRNAs in the sample, we wanted to examine how well low numbers of mapped reads represented the sample when compared with a very large number of mapped reads. In other words, how well do a small subset of the reads represent the distribution of miRNAs in the sample. This analysis uses mapped reads whereas the previous analysis used post-clipped reads. Using a plasma sample, we began with 100,000 randomly chosen mapped reads, and incrementally increased the number of reads we included by 100,000 up to 3.5 million. We then calculated how well the subset of the lower numbers of read inputs, such as 100,000, correlated with the final total of 3.5 million mapped reads (Figure 6). If the correlation is high with a low number of reads, it suggests that low coverage may be sufficient to represent the sample. Selecting reads randomly (instead of miRNAs) preserves the original distribution of reads from the sequencer. We report the Spearman correlations using the three different tools: miRDeep2, miRNAKey, and miRExpress. The correlation becomes fairly stable at ∼1.5 million random reads, correlation coefficient of 0.97 for miRNAKey. Our main observation was that increasing the reads from 1 to 3 million reads only increased the Spearman correlation by 0.05.


Comparison of Analysis Tools for miRNA High Throughput Sequencing Using Nerve Crush as a Model.

Metpally RP, Nasser S, Malenica I, Courtright A, Carlson E, Ghaffari L, Villa S, Tembe W, Van Keuren-Jensen K - Front Genet (2013)

The correlation of a subset of reads, beginning with 100,000 with the full number of mapped reads, 3.5 million.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3585423&req=5

Figure 6: The correlation of a subset of reads, beginning with 100,000 with the full number of mapped reads, 3.5 million.
Mentions: In addition to examining how the incremental increase in coverage altered the detection of new miRNAs in the sample, we wanted to examine how well low numbers of mapped reads represented the sample when compared with a very large number of mapped reads. In other words, how well do a small subset of the reads represent the distribution of miRNAs in the sample. This analysis uses mapped reads whereas the previous analysis used post-clipped reads. Using a plasma sample, we began with 100,000 randomly chosen mapped reads, and incrementally increased the number of reads we included by 100,000 up to 3.5 million. We then calculated how well the subset of the lower numbers of read inputs, such as 100,000, correlated with the final total of 3.5 million mapped reads (Figure 6). If the correlation is high with a low number of reads, it suggests that low coverage may be sufficient to represent the sample. Selecting reads randomly (instead of miRNAs) preserves the original distribution of reads from the sequencer. We report the Spearman correlations using the three different tools: miRDeep2, miRNAKey, and miRExpress. The correlation becomes fairly stable at ∼1.5 million random reads, correlation coefficient of 0.97 for miRNAKey. Our main observation was that increasing the reads from 1 to 3 million reads only increased the Spearman correlation by 0.05.

Bottom Line: We also found that samples with very low starting amounts of RNA (mouse plasma) made high depth of mature miRNA coverage more difficult to obtain.We explored the changes in total mapped reads and differential expression results generated by three different software packages: miRDeep2, miRNAKey, and miRExpress and two different analysis packages, DESeq and EdgeR.We also examine the accuracy of using miRDeep2 to predict novel miRNAs and subsequently detect them in the samples using qRT-PCR.

View Article: PubMed Central - PubMed

Affiliation: Collaborative Bioinformatics Center, Translational Genomics Research Institute Phoenix, AZ, USA.

ABSTRACT
Recent advances in sample preparation and analysis for next generation sequencing have made it possible to profile and discover new miRNAs in a high throughput manner. In the case of neurological disease and injury, these types of experiments have been more limited. Possibly because tissues such as the brain and spinal cord are inaccessible for direct sampling in living patients, and indirect sampling of blood and cerebrospinal fluid are affected by low amounts of RNA. We used a mouse model to examine changes in miRNA expression in response to acute nerve crush. We assayed miRNA from both muscle tissue and blood plasma. We examined how the depth of coverage (the number of mapped reads) changed the number of detectable miRNAs in each sample type. We also found that samples with very low starting amounts of RNA (mouse plasma) made high depth of mature miRNA coverage more difficult to obtain. Each tissue must be assessed independently for the depth of coverage required to adequately power detection of differential expression, weighed against the cost of sequencing that sample to the adequate depth. We explored the changes in total mapped reads and differential expression results generated by three different software packages: miRDeep2, miRNAKey, and miRExpress and two different analysis packages, DESeq and EdgeR. We also examine the accuracy of using miRDeep2 to predict novel miRNAs and subsequently detect them in the samples using qRT-PCR.

No MeSH data available.


Related in: MedlinePlus