iHMMune-align: application of HMM to immunoglobulin partitioning
A complete description of the iHMMune-align algorithm has been published (1). A hidden Markov model (HMM) consists of states and states can be linked through transitions. A transition from one state to another carries a given probability. There are three main types of states within the HMM for the human heavy chain immunoglobulin sequences. One type of state describes the nucleotides within a V, D or J gene. Another describes the N regions and there are states for the exonuclease removals. The final state types are linking states that are used to string the model together, marking the start and ends of various regions, but do not acutally form part of the rearranged sequence.Partitioning using the HMM is achieved by attempting to allocate each nucleotide within an input sequence to the states described above. The states are linked by transistions which have probabilities associated with them. The transitions essentailly create rules for which states can link to other states, while the probabilities give the likelihood that one state does move to another. The path through the states that gives the highest probability gives the most likely partitioning of the sequence.
For example, if we consider a nucleotide (Vn) towards the end of the V-REGION, does the next nucleotide in the sequence (Ntn+1) also belong to the V-REGION?. This nucleotide could potentially be part of the IGHV gene (a V state), part of the N region or even part of the D-REGION. The linking states are used to mark the end of the V-REGION, the start and end of the N1 regions and the start of the IGHD. When considering where to allocate the Vn+1 nucleotide the tranistion probabilities are used. There will be a probabilitiy for the Vn+1 nucleotide being part of the V-REGION, moving to the exonuclease state and hence potentially being part of the N-REGION or D-REGION.
References
1. B. A. Gaeta, Malming, H. R., Jackson, K.J.L., Bain, M.E., Wilson, P., Collins, A. M.,
2007, iHMMune-align: Hidden Markov model-based alignment and identification of germline genes in rearranged immunoglobulin gene sequences ,
Bioinformatics, doi: 10.1093/bioinformatics/btm147
Help
Further Information
Contact
For queries about iHMMune-align:
Andrew Collins
Katherine Jackson