Contact: Tyler Alioto (tyler.alioto@crg.es)

Back to the course homepage: /courses/IRB-INB07/

In this section we use several gene prediction programs on a particular genomic DNA sequence. For each of these programs we obtain a prediction of a candidate gene and we will analyze the differences between predictions and the annotation of the real gene.

The programs we are going to use are geneid and augustus which are available through a web interface.(If we had more time, we would also run genscan, fgenes, or as many gene predictors as we could...) In these, and in many other tools in the web, we access a form where we can paste, or submit, the sequence we want to analyze, and then we press a button in the form that starts the computing process in some computer where the program runs. Once this process is finished, we get a new page in our browser with the results, which in this case should be a predicted gene.

We are going to work with the sequence HS307871, which is stored in FASTA format. This sequence contains one gene, annotated in the following EMBL and NCBI records. Try to identify in these records the different pieces of information related to the annotation of the gene.

In order to use geneid follow these steps:

1. Connect to the geneid server by going to /software/geneid/geneid.html.
   
   
2. Paste the DNA sequence.
   
   
3. Select organism (human).
   
   
4. Run geneid with different (output) parameters:
   • Searching signals: Select acceptors, donors, start and stop codons. For each type of signal, try to find the real ones.
   • Searching exons: Select All exons and try to find the real ones.
   • Finding genes: You do not need to select any option (default behavior). (Description of output formats)
     
     
5. Compare the prediction with the real annotation.
   • By taking a look to the graphical representation of the predicted sites and exons.
   • By inspection of the output and the EMBL/NCBI record.
   • By taking a look to the graphical representation of both, the output and the EMBL/NCBI annotation in this link.
     
     

Please read the GeneID documentation (/software/geneid/docs/) if you need further clarification regarding input and output of geneid.

In order to use augustus follow these steps:

1. Connect to the augustus server by going to http://augustus.gobics.de/submission.
   
   
2. Paste the DNA sequence.
   
   
3. Select species (Homo sapiens).
   
   
4. Compare the prediction with the real annotation.
   • By inspection of the output and the EMBL/NCBI record.
   • By taking a look to the graphical representation of both, the output and the EMBL/NCBI annotation in this link.
     
     

Improve the prediction from some confirmed evidence.

1. Below the text box where we pasted the DNA sequence, we can find a text box where we can paste evidences, which should consist of one or more exons (in GFF format) that are, e.g., experimentally confirmed.
2. In this case we are going to paste as evidence the second exon which has not been predicted correctly by geneid. For the geneid server, select and copy the GFF line corresponding to this exon contained in this file. For the augustus server, select and copy the GFF line corresponding to this exon contained in this file.
3. Paste the line into the evidences or constraints that anchor... text boxes for geneid and augustus, respectively, and run again geneid and augustus on the sequence.
4. For augustus, try forcing it to make a complete gene prediction.
5. Compare the results with the real annotation. What has changed from the previous prediction?
   
   
6. If you have time, try forcing the first/initial exon. Here are the relevant gff files for geneid and augustus.
   
   

We can visualize all gene predictions made using the gff2ps webserver (http://bioweb.pasteur.fr/seqanal/interfaces/gff2ps.html). Upload the gff files for each of the predictions, as well as the EMBL annotation.

For the impatient, here are the results.

Go to the page where we saw the NCBI record, click on the link CDS, and, next to the Display button, unroll the menu box and select the display option FASTA. Now press the button Display, and we will obtain the protein-coding DNA sequence of this gene in FASTA format. You should now have this sequence.

Select the entire sequence (first line is not necessary) and go to the UCSC genome BLAT search by goint to http://genome.ucsc.edu/cgi-bin/hgBlat. In the big text box, paste the coding sequence we just copied, and press the Submit button on the top-right corner of this page.

The result is a single match, where we find two links, browser and details. Visit first the details link and try to understand the the information provided there. Then go backwards and visit the browser link where we will see where this gene is located within the Human genome, as well as other annotated information as EST spliced alignments, etc.

In order to use genscan follow these steps:

1. Connect to the genscan server by going to http://genes.mit.edu/GENSCAN.html.
   
   
2. Paste the DNA sequence.
   
   
3. Select organism (vertebrate).
   
   
4. Compare the prediction with the real annotation.
   • By inspection of the output and the EMBL/NCBI record.
   • By taking a look to the graphical representation of both, the output and the EMBL/NCBI annotation in this link.
     
     

In order to use fgenesh follow these steps:

1. Connect to the fgenesh server by following this link.
   
   
2. Paste the DNA sequence.
   
   
3. Select organism (human).
   
   
4. Compare the prediction with the real annotation.
   • By inspection of the output and the EMBL/NCBI record.
   • By taking a look to the graphical representation of both, the output and the EMBL/NCBI annotation in this link.