Sam:LabNotes/Microbiome-new/2010-12-8
Jump to navigation
Jump to search
General statiscs (whole reads) and contamination analysis (failed-to-align reads) - HC-MDA-Ecoli-SAGs-Ind1to8[edit]
Background[edit]
- Samples were prepared from -(short and long amplified E.coli SAGs -3rd test 11-07-'10)
- Eight E.coli SAGs samples were made to Illumina sequencing libraries using (N2)Primer-based barcodes ID1 to ID8.
- These libraries were equally pulled (20ng per library) and sequenced in pair-end(PE) mode.
- The sequenced data processed by Dr. Zhang for base-calling and saved into FASTQ files.
- Since I don't have genome-tech server permission, I asked Dinh's help to transfer data into Genemapster server.
Analysis approaches[edit]
Single-end data analysis[edit]
- Whole reads alignment using Bowtie => genome coverage calculate
- Script: Read2coverage-kz-3
- Data resource: HC-MDA-Ecoli-SAGs-Ind1to8 (SE reads data set)
- Template:E.coli-k12-mg1655 template
- Bowtie setting: -k 1 -l 28 -n 2 -m 1 --best --best --phred64-quals
- Downstream: Data was used for genome coverage calculation.
File:Sam120610-wholereadbotie-2.jpg
- Using the same alignment processing, I extract the failed-to-aligned and non-specific reads into different folder.
- Script: Read2coverage-kz4
- Data resource: HC-MDA-Ecoli-SAGs-Ind1to8
- Template:E.coli-k12-mg1655 template
- Bowtie setting: -k 1 -l 28 -n 2 -m 1 --best --best --phred64-quals --un [failed-to-aligned file output] --max [nonspecific file output]
- ABySS de novo assembly of failed-to-align reads with (using single-end assembly with k-mer 27)
- Using abyss-fac function to calculate statistics
ABySS-fac report File:Sam120610-abyss assembly of fail-to-aligne reads.jpg
- Length filtering and keep only >300bp contigs.
- Script: seqLenFilger.pl
- Perform BLAST on filtered contigs
- BLAST setting: -K 1 -m 8 -a 2 -v 1 –b 1(using -m 8 output format to extract GI number)
- BLAST setting: -K 1 -m 8 -a 2 -v 1 –b 1(using -m 9 output format to get BLAST statistics)
- NOTE: Use -K 1 -v 1 -b 1 for reporting only one hit
- Taxonomic analysis by NCBI Batch Entrez (against nucletide database). Transform GI number to UID for taxonomic analysis
Results[edit]
px300 px300 Details of blast hits (PDF) Media:ID2-failed-to-align-L300-accession.pdf Media:ID3-failed-to-align-L300-accession.pdf Media:ID4-failed-to-align-L300-accession.pdf Media:ID5-failed-to-align-L300-accession.pdf Media:ID6-failed-to-align-L300-accession.pdf Media:ID7-failed-to-align-L300-accession.pdf Media:ID8-failed-to-align-L300-accession.pdf
E.coli taxonomic ranks ------------------------- Domain: Eubacteria; Phylum: Proteobacteria; Class: Gammaproteobacteria; Order: Enterobacteriales; Family: Enterobacteriaceae; Genus: Escherichia; Species: E. coli
Yeast taxonomic ranks - (to detect if Yeast DNA contamination was introduced through Yeast tRNA carrier) ------------------------ Super kindom: Eukaryota; Domain: Fungi; Phylum: Ascomycota; Class: Saccharomycetes; Order: Saccharomycetales; Family: Saccharomycetaceae; Genus: Saccharomyces; Species: Saccharomyces cerevisiae
Discussions[edit]
- Most of fail-to-align reads can only be assembled to short contigs (under k-mer 27)
- The contigs assembled from fail-to-aligned reads(against E.coli template) contain a significant amount of human DNA contigs.
ID1 ID2 ID3 ID4 ID5 ID6 ID7 ID8 ----------------------------------------------------------- Homo hits N/A 22 0 25 155 86 35 83 All hits N/A 75 15 106 200 145 56 120 (%) 29.3% 23.6% 77.5% 59.3% 62.5% 69.1%
All failed-to-align contig hit
- A certain level fail-to-align contigs also match E.coli genome (e.g. ID-3 and ID-8). It is unknown why some E.coli reads are failed to map to E.coli genome in bowtie.
- Sequencing error => mis-match >3bp => fail to meet mapping criteria in bowtie?
- Chimeric junction in these "single-end" reads?
- ID-1 showed no BLAST hit using length>300 contigs. I tried using length>200 contigs for BLAST but it still came out no hit.
- I am thinking that maybe it's a better idea to use MEGAN metagenomics tool for analyzing these fail-to-align reads directly, which it can draw a taxonomic distribution plot easily.
Testing using less-stringent report suppression setting on bowtie mapping[edit]
- I noticed that the genome coverage calculation script (Read2coverage-kz-3) calculate the coverage from bowtie.out file, however the bowtie.out file only report unique mapped reads (confirmed on http://genome-tech.ucsd.edu/LabNotes/index.php/Sam:LabNotes/Microbiome-new/2010-12-13).
- I guess there are genome coverages not counted if they are from non-unique mapped reads, since non-uniqe mapped reads were not reported in bowtie.out.
Re-analyze genome coverage using new scripts with less-stringent report suppression[edit]
- test data set: s_4_ID6.txt from SE data set (HL083 E.coli SAGs)
- This data set has the highest genome coverage than others.
- Read2coverage-kz-3-suppress_5
- (allow reporting 5 alignments from 1 read)
- Results
# reads processed: 2372473 # reads with at least one reported alignment: 1512449 (63.75%) # reads that failed to align: 512805 (21.61%) # reads with alignments suppressed due to -m: 347219 (14.64%) Reported 1512449 alignments to 1 output stream(s) total bps coverred = 1824363
- Read2coverage-kz-3-suppress_10
- (allow reporting 10 alignments from 1 read)
- Results
# reads processed: 2372473 # reads with at least one reported alignment: 1857421 (78.29%) # reads that failed to align: 512805 (21.61%) # reads with alignments suppressed due to -m: 2247 (0.09%) Reported 1857421 alignments to 1 output stream(s) total bps coverred = 1869404
Conclusions[edit]
- Test data: E.coli SAG HL083(SE) s_4_ID6.txt
Report suppressor | m = 1 | m = 5 | m= 10 |
proccessed reads | 2372473 | 2372473 | 2372473 |
unique alignment | 1411597 (59.50%) | 1512449 (63.75%) | 1857421 (78.29%) |
failed-to-align | 512805 (21.61%) | 512805 (21.61%) | 512805 (21.61%) |
non-unique alignment | 448071 (18.89%) | 347219 (14.64%) | 2247 (0.09%) |
Total covered bp | 1778296 | 1824363 | 1869404 |
Coverage (E.coli K12) | 38.33% | 39.32% | 40.30% |
- Tune down the report suppression strigency did work on increasing genome coverage, but the effect is limited. Even with m=10 setting where alomst all non-unique alignments (18%) were reported (counted in genome coverage), the genome coverage was only increased by ~2% at most.
- The genome coverage is mainly contributed by unique alignment.
Identify where did those non-unique reads map[edit]
- Perform ABYSS on non-unique read of s_4_ID6.nonspecific reads
- I mapped the assembled contigs on E.coli reference genome using BLAT, and found that these contigs are randomly distributed.