Sam:LabNotes/Microbiome-new/2011-4-19

From ZhangLabWiki
Jump to navigation Jump to search

Recruit metagenomic contigs using mouse SAG contigs as baits - BLAT plus Velvet [edit]

Objective[edit]

  • Extract more contigs from metagenomic data into mouse GI microbiome SAG data.
  • If this approach is working, the increased data amount will help us to extend SAG contig length and build up genome of unknown bacteria.

Background[edit]

  • I did BLATing test and found that BLAT can working either 1 string reference genome or multiple sequence references.
  • Stringency test:
    • I BLAT metagenomic contigs(query) against SAG contigs (reference). I pick up four BLAT results for alignment analysis. Two of queries aligned to reference template at 40~50%, which suggest the default BLAT stringency is suitable for our recruiting purpose.

Procedures[edit]

BLATing metagenomic contigs against mouse SAG contigs[edit]

  • Query: Mouse metagenomic contigs(HC9, HC10, HC11, HC12)- assembled by Dr. Zhang (soapK27M3)
  • Target: Mouse SAGs contigs batch A (ID1 ~ ID12) - assembled by Dr. Zhang (soapK27M3)
Command:
 
samchiang@genemapster:~/Tools/blatSuite34$ ./blat -out=psl /media/disk-2/samchiang/Raw-Read-backup/Bigelow_MouseSAG/kz_SOAP_denovo/MmSAG_BatchA_HL081_s6-8/s6-8_A_ID1.contig.100up /media/disk-2/samchiang/Raw-Read-backup/Bigelow_MouseSAG/kz_SOAP_denovo/Mm_Matagenom_HL087_s5-8/HC10_RLV.Soap.K27M3.contig.100up /media/disk-2/samchiang/Analysis_output/BLAT_Meta2SAG_040311/Reg_BLAT_HC10_to_SAG_A_ID1 &

Fetch FASTA sequence from BLAT output[edit]

  • Using galaxy workflow - Fetch_FASTA_from_BLAT_output (040411)
    • Input Dataset 1: Original mouse metagenomic contigs
    • Input Dataset 2: BLAT output (aligned metetagenomic contigs)

Pool fatched FASTA file with SAG contigs[edit]

  • Pooling is based on the SAG ID
    • e.g four recruited sets (HC9, HC10, HC11, HC12) recruited by SAG_A_ID1 contigs, then all of these five sets of contigs are pooled together.

Joint all of the contigs in pool[edit]

Command: 
samchiang@genemapster:/media/disk-1/samchiang/Sub_tests_folder_imp/2_Extending_Mouse_Bac_SAG_contigs/BLAT_Meta2SAG_040311/Step4_pooled_together/SAG_A_ID1$ cat Reg_BLAT_HC11_to_SAG_A_ID1_11reads.fasta.fasta Reg_BLAT_HC10_to_SAG_A_ID1_1567reads.fasta Reg_BLAT_HC12_to_SAG_A_ID1_945reads.fasta Reg_BLAT_HC9_to_SAG_A_ID1_163reads.fasta.fasta s6-8_A_ID1.contig.100up > SAG_A_ID1_and_recruted_Meta_all

Velvet assembling[edit]

  • Test parameters settings and k-mers:
    • (-short, -long) x (k-mer 21, 24, 27)
  • I found that k-mer 27 gave me the best result. The length setting "-long" won't work for for whole contigs dataset. Maybe these contigs are not long enough to use -long setting.
  • For some super long (>32kbp)contigs (like in ID3 and in ID4), I need to use "make LONGSEQUENCES" before I can run the assembly.
    • For super long contigs, I also tried to separate them and assembly them using mix length command -short [contigs < 32Kbp] -long [contigs > 32kbp]. There is no significant difference compared to use -short for all contigs.
  • Used velvet settings
    • velveth: k-mer 27, -short
    • velvetg: fixed setting as cov_cutoff 2, min-contig lgth 100
Command:

samchiang@genemapster:~/Tools/velvet_1.1.03$ ./velveth /media/disk-1/samchiang/Analysis_out/velvet_test_SAG_A_ID3_K27_short 27 -fasta -short /media/disk-1/samchiang/Sub_tests_folder_imp/2_Extending_Mouse_Bac_SAG_contigs/BLAT_Meta2SAG_040311/Step4_pooled_together/SAG_A_ID3/SAG_A_ID3_and_recruted_Meta_all  

samchiang@genemapster:~/Tools/velvet_1.1.03$ ./velvetg /media/disk-1/samchiang/Analysis_out/velvet_test_SAG_A_ID3_K27_short/ -cov_cutoff 2 -min_contig_lgth 100 &

Results[edit]

  • Using abyss-fac to calculate contigs statistics
 File:041911-velvet-results 1.png

File:041911-velvet-results 2.png

File:041911-velvet-N50-plot.jpg
File:041911-velvet-sum-plot.jpg

Discussion[edit]

  • In general, velvet assembly of SAG contigs with recruited contigs increased the N50 to 10~15 times longer than original SAG contigs length. Also the assembly increased about 10%~40% of total bp than original SAG contigs. However, in ID2, ID9 and ID11, the re-assembled contigs have total sum bps than non-assembled SAG contigs.

Next[edit]

  • It is unknown why some assembly get worse results. Maybe I should try ABYSS on those datasets.
  • Validation: I am thinking to use following method to validate if these new assembled contigs are accurate.
    • Tetramer freq counting -> SOM or PCA analysis -> check if contigs in same library are grouped together.
    • Using taxonomic-based approach (e.g. BLASTN) for validation.