# Hidden Markov Model

I read about the hidden markov model in bioinformatics. I amn't able to understand what it is. Can anyone explain me in brief and in very very simple words what it is ?

I have NO background in bioinformatics or programming and wikipedia article on it is too scary .

-
Markov models are used in almost every scientific field. In short, it is a kind of stochastic (random) model and a hidden markov model is a statistical model where your system is assumed to follow a Markov property for which parameters are unknown. Because it is not specific to biology, I would recommend you to ask your question on crossValidated.SE – Remi.b Mar 4 '14 at 7:49
This question appears to be off-topic because it is about a statistical/computational concept – WYSIWYG Mar 4 '14 at 7:51
<this is about to get closed. I think it could be on topic if you edit it to make it specifically about the applications of HMMs to biology. For example in gene prediction or in the building of sequence profiles (e.g. hmmer) – terdon Mar 7 '14 at 6:17
@terdon.. I guess this problem comes from the way bioinformatics is taught.. When you start introductory bioinfo and read about the different tools and databases you often see some vague statement like- The sequences are discovered using HMM and then a non redundant database is made. The term is heavily used but not explained. Nobody usually uses or are even aware of tools like HMMER at this stage. I have also remained confused about this term called HMM for a while. This is still a relevant question but is out of scope of this forum- that's why closing it. – WYSIWYG Mar 7 '14 at 12:29
@WYSIWYG this is exactly the reason why i asked the question. But i understand why this is out of scope. – biogirl Mar 7 '14 at 12:57

A Markov model is a description of a system that follows a Markov process. In a Markov process the next state of the system is a function of its current state and does not depend on where from and how it started. For example Brownian motion can be called a Markov process. The transition from current state to next state is described by probabilities.

So to conclude, a Markov model is a probabilistic model of a system that is assumed to have no memory.

In a Hidden Markov Model (HMM) the state(s) of the system are not known (therefore hidden). However, there are some functions that depend on the state and their outputs can be used to approximate the state of the system. There goes the definition- it is not that easy to explain how it is applied to different concepts (not just bioinformatics).

Strictly speaking this is off-topic. HMM is not a biological concept. It is a statistical/computational concept (it is like asking what is Dynamic Programming- which is the underlying concept behind BLAST algorithm).

-
Thank you for answering my question though I now understand that it is not strictly on topic. – biogirl Mar 4 '14 at 8:07
I read about this in relation to protein domains and motifs. What does that have to do with HMM ? – biogirl Mar 4 '14 at 8:07
Motif finding algorithms use HMM to predict the correct motif.. – WYSIWYG Mar 4 '14 at 11:52
You can check this paper out to know about how HMM is applied in motif discovery. It is by Sean Eddy who has developed many popular HMM based tools – WYSIWYG Mar 7 '14 at 12:17