Lecture 5 "Fuctional descriptionand annotaton transfer"

the more you sequence, the higher the chance to pick up ypur gene of interest

• sort inear sequences that serve a specific function for the protein, but will not be stable or fold independent of the rest of the chain

• often represented as regular expression, for example, C-x-(C)-(DN)-x(5)-C-C

• protein-interaction, ligand interactions, cleavage sites, targeting

• often used are upstream sequences in the transcription binding site

• from prior experiments we know ROX1binding sites, they form training data

• created out of training data

• four lines

• counts the of the number of characters of the seqquenxe

• with the Kulback-Leibler Distanz the distribution is measured

• if you change letters n position of the sequence the score changes

Measures the difference between two probability distribuions

1. Distribution: observed nucleotide frequency at Position i (f b,i)

2. Distribution: expected nucletoide frequency at position i given the overall (position-independet) nucleotide composition of te sequences

• represent extrinsic information in he analysis of protein sequences

• typically short sequence stretches that are linked to a particular function

• Represented by regular expression or position specific scoring matrices (PSSM)

• PSSMs typically contain prior information to account for non-observed data

• fixed length, i.e. gap-free

• motif identification suffers from a high flse positive rate. Reason: short motifs have a hogh chance of occurring at random

• often used for the prediction of bidning sites, e.g. transcription factor binding sites, and for the prediction of modification sites

• PSSms can also be used for FAST omain annotation

1. each column in an algnment is reprsented by one column in the PSSM, and by 0 to 1 states is a pHMM

2. The occurrence of each letter in the alphabet of the analysed sequence type is associated with a colmn-specific score in the PSSM, an with an emission porbability in the pHMM

3. The order of columns in the pHMM is fixed and there is no alternative from visiting the columns from left to right. In pHMMs there is a transition probability <1 between connected states

• position in the lignment for which all or at least most of the sequences are represented by amino acid. In evolutionary terms, a match state is an ancestral position with a very low deletion probability

• the idea is to generate a HMM with a repetitive strucutre of states that differ in their emission probabilities for the 20 aa

• D=deletion states

• M= matching states

• I = insertion state

• catches loops singe sequences are variable

• catches variabilities with models wit same function

• represent extrinsic information in the analysis of protein sequences

• model typically longer sequence stretches that are linked to a particular function, or are simply evolutionary conserved (c.f. Domain od unknown function (DUF) in Pfam)

• contain priot information to account for non observed data

• variable length, i.e. explicit modeling of insertion and deletions

• typically highly specific and allow the identification even in highly diverges sequences

• commonly used for the prediction of functional domains

• can be arranged in domain architectures capturing the context in which individual domains occur

• RRM = RNA recognition motif

  • subcellular localisation

  • ineraction partners

  • physiological reference - involved in retinal function

• not indvidual protein domains, but the complete feature architecture (ie. the combination of multiple protein domains and features) defines the function of a protein)

How to describe protein function - or more generally biological knowledge- in a way that makes it accessible to automated analysis?

Controlled vocabulary

• definition: arranged list of standardizing terminology

• takes the guess work out of searching

• allows categorization

• example: yellow pages

• Different level of complexity:

  • equivalence relationships

  • hierachical relationships (taxonomy)

  • associative relationships (thesaurus)

• describes the localisation within the cell, anatomical strucute of macro-molecular complex (e.g. ribosome)

describes the molecular activities, such as catalytic or binding activity (e.g. kinase) does not specify the context of action

describes a series of events in a physiological process, e.g. signal transduction, not a pathway

• parent & child terms

• ìsa`, `part_of` and `regulates`relationships

• directed acylic graphs

• more flexible than hierachy

• each gene is annotated with set of most specific terms that describe its functionality

• transcriptional profiling (microarray/RNA-seq) to compare condition, e.g. healthy vs. disease tissue

• search for terms that are significantly overrepresented

• developed to describe gene function such it accesible to computer-aided analysis

• based on a restricted vocabulary (describing entities are referred to as `terms`)

• terms are hierchally organized and are `related`to each other

• term of relationships are organized in a directed acylic graph

• GO term - gene association is based on different evidences

• evidence code inferred by electronic annotation is better than expected

• GO term enrichment analyses help to identify terms that occur (or less) ften than expected by chance

• Graphical fisplay of GO enrichment analysis otcome helps interpreting the result

• GO slim, when not all information in the GO is relevant