Compared with inward paired-end sequencing data, the outward sequencing data from mate-pair reads maybe contain the inside adapter sequence, due to the random shearing. So how can I efficiently align these mate-paired reads? Does anyone have such experience in aligning or de novo assembly? Thanks a lot.
The answer depends on the kind of data you have and the assembler you use.
Our standard pipeline is based on the Nextera MatePair kit for Illumina sequencers. We size select the library in a way that we end up with read pairs (2x 250-300bp on a MiSeq) that overlap by ~50bp. The pairs are then joined on that overlap, discarding all pairs that cannot be joined. The joined reads are then searched for the presence of the internal adapter, discarding all where it cannot be found. The reads are split on the adapter and reformatted so that they appear to be paired reads from a large insert library. These are then assembled with the Newbler assembler. Works like a charm for bacterial genomes.
For HiSeq data from a collaborator (2x 100bp from 600-700bp fragments, so no overlap) I used a different approach. There, all reads containing the complete linker were discarded (as the amount of data lost was negligible), the remaining reads were reverse complemented (to appear like standard paired reads from a large insert library) and Newbler was run with the linker sequence used for vector trimming (to remove linker traces at the end).
Hope that helps a bit.
The answer depends on the kind of data you have and the assembler you use.
Our standard pipeline is based on the Nextera MatePair kit for Illumina sequencers. We size select the library in a way that we end up with read pairs (2x 250-300bp on a MiSeq) that overlap by ~50bp. The pairs are then joined on that overlap, discarding all pairs that cannot be joined. The joined reads are then searched for the presence of the internal adapter, discarding all where it cannot be found. The reads are split on the adapter and reformatted so that they appear to be paired reads from a large insert library. These are then assembled with the Newbler assembler. Works like a charm for bacterial genomes.
For HiSeq data from a collaborator (2x 100bp from 600-700bp fragments, so no overlap) I used a different approach. There, all reads containing the complete linker were discarded (as the amount of data lost was negligible), the remaining reads were reverse complemented (to appear like standard paired reads from a large insert library) and Newbler was run with the linker sequence used for vector trimming (to remove linker traces at the end).
Hope that helps a bit.
Christan's answer is spot-on: there are different approaches, but in most cases the internal adapter sequence is removed by splitting the reads at the adapter and then trimming it away before alignment.
If I am not mistaken there is one exception: in the software that is supplied with the Roche 454 systems you can specify the internal adapter and it is removed on the fly during alignment/assembly - but this software only works with 454 raw data.
Nicholas is right. For 454 data the Newbler assembler itself takes care of everything. This includes filtering out short reads (< 30bp if I remember correctly) post splitting.
And due to the long reads the number of partial adapters at the ends is fairly limited. So even if they get past the filtering steps, there are usually not enough of them to break the assembly process.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools