Hi David, I'm sorry for a slow response. Relatively recently I solved a problem a bit like this and would be happy to share more information with you. If your genome is small I think it makes sense to map to a reference and identify variant sites. (In my opinion de novo assembly isn't needed - see below).
A basic approach is: groom FASTA file -> map with BWA -> filter SAM (uniquely mapped reads only) -> SAM-to-BAM -> Generate pileup -> Filter pileup
This gives you a position-by-position summary relative to the reference. And that last step is important and needs the most care: you can have it print out differences total numbers of non-reference bases. I can share some information about thresholding how many of these constitute significant evidence that a non-reference base is actually there at that position (basically I use a binomial distribution and ask whether the distribution of ref/non-ref would occur by chance). Given that coverage of small genomes tends to be high, your first question about determining the actual genome sequence (or the quasispecies consensus if you prefer!) can be answered by majority rules: i.e., a small script (or with tools under "Text Manipulation" heading) to read off the base with the most support at each position and then to test whether that base == base in reference nucleotide column.
It's probably also worth thinking about PCR duplicates (from library prep) as these could be a significant source of error, but they are also tricky when many reads will be identical anyway in the input DNA.
Feel free to get in touch with me if you need a bit more clarity and/or some more specifics...
cheers, Ben
On Apr 4, 2011, at 9:55 PM, Anton Nekrutenko wrote:
From: David Matthews D.A.Matthews@bristol.ac.uk Date: April 4, 2011 6:02:03 PM EDT To: galaxy-user@lists.bx.psu.edu Subject: [galaxy-user] Assemble a consensus genome from NGS data
Hi,
Does anyone know how to get a consensus genome from NGS data indicating the percent variance at each nucleotide? I have a small virus genome with manyfold coverage from my transcriptomic run. I'd like to know what the transcriptome indicates is the actual genome plus get a feel for any hotspots where there appears to be significant varience from the reference sequence (i.e. because the reference is wrong or perhaps because of frequent errors in that region due to RNA pol II having a problem accurately transcribing the sequence).
Many thanks!
David
The Galaxy User list should be used for the discussion of Galaxy analysis and other features on the public server at usegalaxy.org. Please keep all replies on the list by using "reply all" in your mail client. For discussion of local Galaxy instances and the Galaxy source code, please use the Galaxy Development list:
http://lists.bx.psu.edu/listinfo/galaxy-dev
To manage your subscriptions to this and other Galaxy lists, please use the interface at:
Benjamin Dickins Postdoctoral Researcher Center for Comparative Genomics and Bioinformatics The Pennsylvania State University ------------------------------------------------------------ 302 Wartik Laboratory University Park, PA 16802, USA Cell/mobile: +1 814 777 1852 Office tel: +1 814 863 2185 Office fax: +1 814 865 9131 Website: http://www.bendickins.net/ Weblog: http://www.open.ac.uk/blogs/ideasblog/
HI Ben,
Do not apologise, this is excellent guidance! I have been bumbling about with pile up and your explanation makes it much clearer. I did not use BWA but tophat instead so I'll give it a go with bwa and see if it makes a difference. I'm off to a virology conference next week so I'm not sure how much chance I'll get to work on it but many thanks again and once I do get my teeth into it I'm sure I'll have some more questions - especially on the stats front. On a related subject I am also looking at indels to see if the virus has hotspots for transcription errors that may reflect a deliberate attempt by the virus to modulate RNApolII function through secondary RNA structure interfering with polII fidelity (I have no other evidence for this, just a mad shot in the dark!). Have you ever looked at this either?
Best Wishes, David
On 8 Apr 2011, at 05:08, Benjamin Dickins wrote:
Hi David, I'm sorry for a slow response. Relatively recently I solved a problem a bit like this and would be happy to share more information with you. If your genome is small I think it makes sense to map to a reference and identify variant sites. (In my opinion de novo assembly isn't needed - see below).
A basic approach is: groom FASTA file -> map with BWA -> filter SAM (uniquely mapped reads only) -> SAM-to-BAM -> Generate pileup -> Filter pileup
This gives you a position-by-position summary relative to the reference. And that last step is important and needs the most care: you can have it print out differences total numbers of non-reference bases. I can share some information about thresholding how many of these constitute significant evidence that a non-reference base is actually there at that position (basically I use a binomial distribution and ask whether the distribution of ref/non-ref would occur by chance). Given that coverage of small genomes tends to be high, your first question about determining the actual genome sequence (or the quasispecies consensus if you prefer!) can be answered by majority rules: i.e., a small script (or with tools under "Text Manipulation" heading) to read off the base with the most support at each position and then to test whether that base == base in reference nucleotide column.
It's probably also worth thinking about PCR duplicates (from library prep) as these could be a significant source of error, but they are also tricky when many reads will be identical anyway in the input DNA.
Feel free to get in touch with me if you need a bit more clarity and/or some more specifics...
cheers, Ben
On Apr 4, 2011, at 9:55 PM, Anton Nekrutenko wrote:
From: David Matthews D.A.Matthews@bristol.ac.uk Date: April 4, 2011 6:02:03 PM EDT To: galaxy-user@lists.bx.psu.edu Subject: [galaxy-user] Assemble a consensus genome from NGS data
Hi,
Does anyone know how to get a consensus genome from NGS data indicating the percent variance at each nucleotide? I have a small virus genome with manyfold coverage from my transcriptomic run. I'd like to know what the transcriptome indicates is the actual genome plus get a feel for any hotspots where there appears to be significant varience from the reference sequence (i.e. because the reference is wrong or perhaps because of frequent errors in that region due to RNA pol II having a problem accurately transcribing the sequence).
Many thanks!
David
The Galaxy User list should be used for the discussion of Galaxy analysis and other features on the public server at usegalaxy.org. Please keep all replies on the list by using "reply all" in your mail client. For discussion of local Galaxy instances and the Galaxy source code, please use the Galaxy Development list:
http://lists.bx.psu.edu/listinfo/galaxy-dev
To manage your subscriptions to this and other Galaxy lists, please use the interface at:
Benjamin Dickins Postdoctoral Researcher Center for Comparative Genomics and Bioinformatics The Pennsylvania State University
302 Wartik Laboratory University Park, PA 16802, USA Cell/mobile: +1 814 777 1852 Office tel: +1 814 863 2185 Office fax: +1 814 865 9131 Website: http://www.bendickins.net/ Weblog: http://www.open.ac.uk/blogs/ideasblog/
Hi David, Thanks! Your indel/pol point sounds very interesting - I've not thought of that before - and I'd be happy to talk indels a bit as I ought to be thinking about these a bit more (and because they can be mistaken for substitutions if one isn't careful)... thanks, Ben
On Apr 8, 2011, at 2:55 PM, David Matthews wrote:
HI Ben,
Do not apologise, this is excellent guidance! I have been bumbling about with pile up and your explanation makes it much clearer. I did not use BWA but tophat instead so I'll give it a go with bwa and see if it makes a difference. I'm off to a virology conference next week so I'm not sure how much chance I'll get to work on it but many thanks again and once I do get my teeth into it I'm sure I'll have some more questions - especially on the stats front. On a related subject I am also looking at indels to see if the virus has hotspots for transcription errors that may reflect a deliberate attempt by the virus to modulate RNApolII function through secondary RNA structure interfering with polII fidelity (I have no other evidence for this, just a mad shot in the dark!). Have you ever looked at this either?
Best Wishes, David
On 8 Apr 2011, at 05:08, Benjamin Dickins wrote:
Hi David, I'm sorry for a slow response. Relatively recently I solved a problem a bit like this and would be happy to share more information with you. If your genome is small I think it makes sense to map to a reference and identify variant sites. (In my opinion de novo assembly isn't needed - see below).
A basic approach is: groom FASTA file -> map with BWA -> filter SAM (uniquely mapped reads only) -> SAM-to-BAM -> Generate pileup -> Filter pileup
This gives you a position-by-position summary relative to the reference. And that last step is important and needs the most care: you can have it print out differences total numbers of non-reference bases. I can share some information about thresholding how many of these constitute significant evidence that a non-reference base is actually there at that position (basically I use a binomial distribution and ask whether the distribution of ref/non-ref would occur by chance). Given that coverage of small genomes tends to be high, your first question about determining the actual genome sequence (or the quasispecies consensus if you prefer!) can be answered by majority rules: i.e., a small script (or with tools under "Text Manipulation" heading) to read off the base with the most support at each position and then to test whether that base == base in reference nucleotide column.
It's probably also worth thinking about PCR duplicates (from library prep) as these could be a significant source of error, but they are also tricky when many reads will be identical anyway in the input DNA.
Feel free to get in touch with me if you need a bit more clarity and/or some more specifics...
cheers, Ben
On Apr 4, 2011, at 9:55 PM, Anton Nekrutenko wrote:
From: David Matthews D.A.Matthews@bristol.ac.uk Date: April 4, 2011 6:02:03 PM EDT To: galaxy-user@lists.bx.psu.edu Subject: [galaxy-user] Assemble a consensus genome from NGS data
Hi,
Does anyone know how to get a consensus genome from NGS data indicating the percent variance at each nucleotide? I have a small virus genome with manyfold coverage from my transcriptomic run. I'd like to know what the transcriptome indicates is the actual genome plus get a feel for any hotspots where there appears to be significant varience from the reference sequence (i.e. because the reference is wrong or perhaps because of frequent errors in that region due to RNA pol II having a problem accurately transcribing the sequence).
Many thanks!
David
The Galaxy User list should be used for the discussion of Galaxy analysis and other features on the public server at usegalaxy.org. Please keep all replies on the list by using "reply all" in your mail client. For discussion of local Galaxy instances and the Galaxy source code, please use the Galaxy Development list:
http://lists.bx.psu.edu/listinfo/galaxy-dev
To manage your subscriptions to this and other Galaxy lists, please use the interface at:
Benjamin Dickins Postdoctoral Researcher Center for Comparative Genomics and Bioinformatics The Pennsylvania State University
302 Wartik Laboratory University Park, PA 16802, USA Cell/mobile: +1 814 777 1852 Office tel: +1 814 863 2185 Office fax: +1 814 865 9131 Website: http://www.bendickins.net/ Weblog: http://www.open.ac.uk/blogs/ideasblog/
Benjamin Dickins Postdoctoral Researcher Center for Comparative Genomics and Bioinformatics The Pennsylvania State University ------------------------------------------------------------ 302 Wartik Laboratory University Park, PA 16802, USA Cell/mobile: +1 814 777 1852 Office tel: +1 814 863 2185 Office fax: +1 814 865 9131 Website: http://www.bendickins.net/ Weblog: http://www.open.ac.uk/blogs/ideasblog/
galaxy-user@lists.galaxyproject.org
-
Benjamin Dickins -
David Matthews