Scribd
Upload
45 views1 page

P (X P (Y: 1. Open Areas For Research in Hmms in Biology

The document discusses limitations of using Hidden Markov Models (HMMs) for biological problems. HMMs make a Markov Chain assumption that events are independent, which does not reflect the dependencies among amino acids. Additionally, standard machine learning problems of local maxima and overfitting occur when training HMMs on biological data. Open areas of research include better integrating structural information into HMMs and exploring different model architectures informed by protein structure knowledge rather than only fitting observed data.

Uploaded by

juanpapo
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
45 views1 page

P (X P (Y: 1. Open Areas For Research in Hmms in Biology

The document discusses limitations of using Hidden Markov Models (HMMs) for biological problems. HMMs make a Markov Chain assumption that events are independent, which does not reflect the dependencies among amino acids. Additionally, standard machine learning problems of local maxima and overfitting occur when training HMMs on biological data. Open areas of research include better integrating structural information into HMMs and exploring different model architectures informed by protein structure knowledge rather than only fitting observed data.

Uploaded by

juanpapo
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 1

Model

So, they are unable to capture higher order correlations among amino-acids.

 Markov Chain assumption of independent events

Probabilities of states are supposed to be independent which is not true of biology

Eg, P(y) must be independent of P(x), and vice versa

P(x … P(y

 Standard Machine Learning Problems

In the training problem, we need to watch out for local maxima and so model may not converge to

a truly optimal parameter set for a given training set. Secondly, since the model is only as good as

your training set, this may lead to over-fitting.

1. Open areas for research in HMMs in biology

 Integration of structural information into profile HMMs.

Despite the almost obvious application of using structural information on a member protein family

when one exists to better the parameterization of the HMM, this has been extremely hard to achieve

in practice.

 Model architecture

The architectures of HMMs have largely been chosen to be the simplest architectures that can fit the

observed data. Is this the best architecture to use? Can one use protein structure knowledge to make

better architecture decisions, or, in limited regions, to learn the architecture directly from the data?

Will these implied architectures have implications for our structural understanding?

You might also like