Notice
3.9. Benchmarking the prediction methods
- document 1 document 2 document 3
- niveau 1 niveau 2 niveau 3
Descriptif
It is necessary to underline that gene predictors produce predictions. Predictions mean that you have no guarantees that the coding sequences, the coding regions,the genes you get when applying your algorithm, are true genes, thatis genes which have a biological existence. Only experimental analysiscan confirm or infirm your predictions. Nevertheless it is interesting and also important to be able to evaluate your algorithm, thisis the role of benchmarking. Benchmarking means measuring the capacity of your algorithm to produce good predictions. How can we make thiskind of measurement? We need a reference, an idealreference would be a genome which is well annotated and for whichall of the annotations have been confirmed through experimental results. Unfortunately, there are very few genomes for which we have this experimental confirmation.Even, for example, the E. coli genome, which is a well-knownorganism and well annotated genome, is not the ideal reference. However, it is also interesting to compare prediction algorithm and method between them, to do acompetition, to apply several predictors on the same genomes and to compare the results of this algorithm.
Thème
Documentation
Dans la même collection
-
3.2. A simple algorithm for gene prediction
RechenmannFrançoisBased on the principle we statedin the last session, we will now write in pseudo code a firstalgorithm for locating genes on a bacterial genome. Remember first how this algorithm should work, we first
-
3.5. Making the predictions more reliable
RechenmannFrançoisWe have got a bacterial gene predictor but the way this predictor works is rather crude and if we want to have more reliable results, we have to inject into this algorithmmore biological knowledge. We
-
3.8. Probabilistic methods
RechenmannFrançoisUp to now, to predict our gene,we only rely on the process of searching certain strings or patterns. In order to further improve our gene predictor, the idea is to use, to rely onprobabilistic methods
-
3.3. Searching for start and stop codons
RechenmannFrançoisWe have written an algorithm for finding genes. But you remember that we arestill to write the two functions for finding the next stop codonand the next start codon. Let's see how we can do that. We
-
3.6. Boyer-Moore algorithm
RechenmannFrançoisWe have seen how we can make gene predictions more reliable through searching for all the patterns,all the occurrences of patterns. We have seen, for example, howif we locate the RBS, Ribosome
-
3.1. All genes end on a stop codon
RechenmannFrançoisLast week we studied genes and proteins and so how genes, portions of DNA, are translated into proteins. We also saw the very fast evolutionof the sequencing technology which allows for producing
-
3.10. Gene prediction in eukaryotic genomes
RechenmannFrançoisIf it is possible to have verygood predictions for bacterial genes, it's certainly not the caseyet for eukaryotic genomes. Eukaryotic cells have manydifferences in comparison to prokaryotic cells. You
-
3.4. Predicting all the genes in a sequence
RechenmannFrançoisWe have written an algorithm whichis able to locate potential genes on a sequence but only on one phase because we are looking triplets after triplets. Now remember that the genes maybe located on
-
3.7. Index and suffix trees
RechenmannFrançoisWe have seen with the Boyer-Moore algorithm how we can increase the efficiency of spin searching through the pre-processing of the pattern to be searched. Now we will see that an alternative way of
Avec les mêmes intervenants et intervenantes
-
1.7. DNA walk
RechenmannFrançoisWe will now design a more graphical algorithm which is called "the DNA walk". We shall see what does it mean "DNA walk". Walk on to DNA. Something like that, yes. But first, just have a look again at
-
2.6. Algorithms + data structures = programs
RechenmannFrançoisBy writing the Lookup GeneticCode Function, we completed our translation algorithm. So we may ask the question about the algorithm, does it terminate? Andthe answer is yes, obviously. Is it pertinent,
-
3.3. Searching for start and stop codons
RechenmannFrançoisWe have written an algorithm for finding genes. But you remember that we arestill to write the two functions for finding the next stop codonand the next start codon. Let's see how we can do that. We
-
4.7. Alignment costs
RechenmannFrançoisWe have seen how we can compute the cost of the path ending on the last node of our grid if we know the cost of the sub-path ending on the three adjacent nodes. It is time now to see more deeply why
-
4.9. Recursion can be avoided: an iterative version
RechenmannFrançoisWe have written a recursive function to compute the optimal path that is an optimal alignment between two sequences. Here all the examples I gave were onDNA sequences, four letter alphabet. OK. The
-
1.2. At the heart of the cell: the DNA macromolecule
RechenmannFrançoisDuring the last session, we saw how at the heart of the cell there's DNA in the nucleus, sometimes of cells, or directly in the cytoplasm of the bacteria. The DNA is what we call a macromolecule, that
-
1.10. Overlapping sliding window
RechenmannFrançoisWe have made some drawings along a genomic sequence. And we have seen that although the algorithm is quite simple, even if some points of the algorithmare bit trickier than the others, we were able to
-
2.3. The genetic code
RechenmannFrançoisGenes code for proteins. What is the correspondence betweenthe genes, DNA sequences, and the structure of proteins? The correspondence isthe genetic code. Proteins have indeedsequences of amino acids.
-
3.6. Boyer-Moore algorithm
RechenmannFrançoisWe have seen how we can make gene predictions more reliable through searching for all the patterns,all the occurrences of patterns. We have seen, for example, howif we locate the RBS, Ribosome
-
4.4. Aligning sequences is an optimization problem
RechenmannFrançoisWe have seen a nice and a quitesimple solution for measuring the similarity between two sequences. It relied on the so-called hammingdistance that is counting the number of differencesbetween two
-
5.2. The tree, an abstract object
RechenmannFrançoisWhen we speak of trees, of species,of phylogenetic trees, of course, it's a metaphoric view of a real tree. Our trees are abstract objects. Here is a tree and the different components of this tree.
-
1.5. Counting nucleotides
RechenmannFrançoisIn this session, don't panic. We will design our first algorithm. This algorithm is forcounting nucleotides. The idea here is that as an input,you have a sequence of nucleotides, of bases, of letters,