BLAT sequence versus different versions of same genome

[jimsac2016]

I used the UCSC BLAT tool to align a canine EST (DN435215, 565 bases in length) to the current version of the dog genome (canfam3). Only the first 175 bases align to the genome. The rest do not align. The ENTIRE EST aligns well with other mammalian reference sequences corresponding to the TRMP1 gene. The entire EST aligns well with the PREVIOUS version of the dog genome (canFam2), as well as with SRA sequences from different dog breeds. Does this mean that the present version of the canine genome is incorrect? I would really appreciate any illumination on this perplexing problem

[Richard Finney]

I think your suspecting a "bad "reference genome spot may be correct.

[jimsac2016]

Problem with canfam3

I think your suspecting a "bad "reference genome spot may be correct.

Can I report this to anybody? The Broad Institute perhaps?

[GenoMax]

Are you sure about this? See a blastn search against the Canis lupus genome from refseq_genome v.3.1 (this link will expire in a day or two): http://blast.ncbi.nlm.nih.gov/Blast....ID=K4PHV905015

[jimsac2016]

Only the first part of the transcript aligns well

Are you sure about this? See a blastn search against the Canis lupus genome from refseq_genome v.3.1 (this link will expire in a day or two): http://blast.ncbi.nlm.nih.gov/Blast....ID=K4PHV905015

First of all, thanks for your interest. Your insight is much valued
Look at how the best alignments only comprise the first 172 bp of the sequence (to chromosome 2), where the putative TRMP1 gene is located. Other parts of the sequence align (less well) to other chromosomes in the dog genome. But this does not happen with the previous version of the canine genome, canfam2, (available at the UCSC website). What is happening here?

[Markiyan]

Different assemblies have some of the problems in different places...

Basically the same region may be assembled well in one assembly and broken in the another one.

This depends on many factors (input DNA, library preparation, sequencing technology(es) used (esp read length/error rate), reads preprocessing, assembler program/settings used (See assemblathron I/II papers for more in-depth discussion).

You can try locating the assembly version where this region is intact and use it or (if you have a lot of bioinformatics resources/experience):

In any case I recommend downloading entire SRA RAW reads dataset(s), preprocessing the reads (trimming the adapters if first gen NGS) and making a BLAST or BLAT database from it.

Than you can use your probe DNA sequence to pull out the reads from the region(s) of interest. Than you can use those reads for miniassembly, which you can edit/fix by the means of consed or similar assembly editor.

Trying to do above can make a good part II/III bioinformatcs student project

[jimsac2016]

Basically the same region may be assembled well in one assembly and broken in the another one.

This depends on many factors (input DNA, library preparation, sequencing technology(es) used (esp read length/error rate), reads preprocessing, assembler program/settings used (See assemblathron I/II papers for more in-depth discussion).

You can try locating the assembly version where this region is intact and use it or (if you have a lot of bioinformatics resources/experience):

In any case I recommend downloading entire SRA RAW reads dataset(s), preprocessing the reads (trimming the adapters if first gen NGS) and making a BLAST or BLAT database from it.

Than you can use your probe DNA sequence to pull out the reads from the region(s) of interest. Than you can use those reads for miniassembly, which you can edit/fix by the means of consed or similar assembly editor.

Trying to do above can make a good part II/III bioinformatcs student project

Wow. That is so fascinating! I am presently using SRAs and the older assembly, together with partial canine ESTs and complete experimentally verified and software predicted full transcripts. I am not a bioinformaticist by training but from your comments it seems I am heading in the right direction with solving this dilemma. What software would you recommend for SRA analysis? Thanks!

[Markiyan]

Start SRA analysis with fastq-dump...

When dealing with SRA - use something like fastq-dump:
(have a look for more info):

https://edwards.sdsu.edu/research/fastq-dump/

Make sure to get the original read id's (-F parameter).
Make one dump in the fastq format.

Than you can check the fastq file with something like fastqc.

If there are adapters/contaminants use something like trimmomatic or prinseq to clean them up.

If reads are paired - feed them into the FLASH.

Than you can convert fastq files to fasta with prinseq,

If no adapters are present - than make a blast database using formatdb from the fasta file. Make sure you know how to extract template name (for paired reads) – ex: for illumina - discard /1 or /2.

(if using legacy ncbi blast), make sure you have less than 100 volumes - subdivide the input fastq if more, than you can join them by the means of *.nal files (NCBI db aliases)

Than you can finally split yours query onto 1-2kb chunks and search for reads in that region, if using first gen NGS: up the wordsize to 32, turn on megablast, lower the expect threshold to 1e-20 or so, and use tabular output -m 8:

blastall -W 32 -e 1e-20 -n T -m 8

Than combine the realists (can use cat *.blast | cut -f 2 | uniq commands) for all chunks and use a perl script to:
1. read the names of all reads to include and extract template names for each of them.
2. add those template names to hash array.
3. Open the output fastq file.
4. Parse input input fastq file used to make the db, get the template name of each read and check whether it is in the include list (hash array).
If yes - write it to the output fastq file.

Than use the reads subset fastq file in yours favorite de novo assembler or mapper...