Chris:LabNotes/Microbiome/2011/2011-5-2
Jump to navigation
Jump to search
Slide Shortening Assembly Test[edit]
Background[edit]
- Sam tried to increase SAG contig length by using recruited sequences from metagenomic contigs (http://genome-tech.ucsd.edu/LabNotes/index.php/Sam:LabNotes/Microbiome-new/2011-4-19)
- But results showed some decreased contig lengths after recruitment (ID2, ID9, ID11)
- Dr. Zhang suggested that the decreased contig lengths may be fixed by slide shortening of contigs
- On 4/29, I wrote perl script that could help perform the slide shortening of contigs
- Here, I test the script/method of slide shortening to increase the contig length after recruitment
- SAG sequence: ID9 because showed the lower contig length but showed relatively good contig length
- Sequences found in: </media/disk-1/samchiang/Sub_tests_folder_imp/2_Extending_Mouse_Bac_SAG_contigs/BLAT_Meta2SAG_040311/Step4_pooled_together/SAG_A_ID9>
- Test found in: </media/disk-2/cjwei/slide-shortening_Assembly_Test>
- Used various sliding cut lengths:
- Test 1: >1000bp contigs cut; step size = 200bp
- Test 2: >2000bp contigs cut; step size = 100bp (same as "Whole Genome Amplification and Denovo Assembly of Single Bacterial Cells")
- Test 3: >1000bp contigs cut; step size = 100bp
- Test 4: >100bp contigs cut; step size = 10bp (most highly cut option)
- Test 5: No sliding cut but have assembly
- Test 6: Control (no sliding cut, no assembly -> just run abyss-fac on the combined file)
- Test 7: SAG contigs (for comparison to determine if recruitment helped)
- Also tried out the sliding cut on another assembly (see if can improve the average recruited assembly results) = ID1
- Test 1: >2000bp contigs cut; step size = 100bp (was shown with ID9 that this produced highest N50 and highest sum)
- Test 2: No sliding cut but have assembly
- Test 3: Control (no sliding cut, no assembly -> just run abyss-fact on the combined file)
- Test 4: SAG contigs (for comparison to determine if recruitment helped)
Procedure[edit]
- Use Sliding Cut for combined SAG and recruited metagenome contigs:
perl contig_sliding_cut.pl SAG_A_ID9_and_recruted_Meta_all >SAG_A_ID9_and_recruted_Meta_all_slide_cut
- Run Velveth
- k-mer 27, -short
/media/disk-2/cjwei/Velvet/velveth SAG_A_ID9_and_recruted_Meta_all_slide_cut_Assembly/ 27 -fasta -short SAG_A_ID9_and_recruted_Meta_all_slide_cut
- Run Velvetg
- fixed setting as cov_cutoff 2, min-contig lgth 100
/media/disk-2/cjwei/Velvet/velvetg SAG_A_ID9_and_recruted_Meta_all_slide_cut_Assembly/ -cov_cutoff 2 -min_contig_lgth 100
Results[edit]
ID9 (problem assembly)[edit]
n n:100 n:N50 min median mean N50 max sum 7871 7871 737 100 196 465 1000 20981 3662057 contigs_1000_200.fa 7819 7819 710 100 197 473 1057 21081 3703868 contigs_2000_100.fa 7861 7861 726 100 196 469 1039 21081 3689257 contigs_1000_100.fa 7869 7869 718 100 196 468 1041 21091 3690448 contigs_100_10.fa 990 990 101 100 211 505 980 19794 500631 SAG_A_ID9_and_recruted_Meta_all_Assembly 14828 14828 1140 100 170 403 902 38317 5988190 ../SAG_A_ID9_and_recruted_Meta_all 392 4392 490 100 187 400 769 9214 1758616 s6-8_A_ID9.contig.100up
- Based on results, notice that >2000bp contigs cut; step size = 100bp had highest N50 and sum
ID1 (average assembly)[edit]
n n:100 n:N50 min median mean N50 max sum 2035 2035 122 100 140 476 1695 17800 968883 contigs_2000_100.fa 342 342 31 100 368 1337 3721 19794 457301 SAG_A_ID1_and_recruted_Meta_all_Assembly 4428 4428 232 100 188 829 3185 19794 3674310 ../SAG_A_ID1_and_recruted_Meta_all 1742 1742 362 100 123 199 216 3710 348324 s6-8_A_ID1.contig.100up
Conclusions[edit]
- Based on ID9 test, >2000bp contigs cut and step size 100bp had best results when using sliding cut method
- When comparing sliding cut method and just assembly of recruited sequences and SAG contigs, sliding cut shows ~2-6x increase in sum, and approximately same (sometimes slightly lower) N50 and max
- Sliding cut method improves recruitment assembly when compared with assembly w/o sliding cut
- Based on both tests, when using sliding cut, recruitment showed a marked increase in assembly (i.e. higher N50, max, and sum) than original SAG contigs -> shows that recruitment helps with assembly (need to validate if assembly is accurate)
- Recruitment w/ sliding cut shows improvement compared to SAG
- However, when comparing sliding cut assembly and just statistics of SAG and recruited contigs (without assembly) = sliding cut shows lower sum/N50/max
- Not expected to beat out combination without assembly because there is redundancy in sequencies
Follow-Up[edit]
- Sam brought out the question: would the number of copies change the sum?
- Here, test whether the number of copies changes any of the contig statistics
- Procedure: Above, the script had 3 copies of each contig (both split and un-split)
- 1. Change script so have 1 copy
- 2. Change script so have 5 copies
- 3. Run both changed scripts on ID9 with >2000bp contigs cut and step size 100bp
- Results
n n:100 n:N50 min median mean N50 max sum 1280 1280 138 100 341 1334 3500 20981 1707966 contigs_2000_100_1_copy.fa 7819 7819 710 100 197 473 1057 21081 3703868 contigs_2000_100_3_copy.fa 7820 7820 710 100 197 473 1057 21081 3704053 contigs_2000_100_5_copy.fa
- Conclusion
- Increasing the number of copies above 3 copies only slightly changes the sum -> the sum plateaus with change of number of copies (but max and N50 are max at number of copies = 3 already)