Thursday, August 29, 2013

Father's or Mother's gene. Which one will be expressed?




It has been long thought that (As in Mendelian genetics) parental origin of gene does not affect its expression. With respect to autosomal genes, male and female contribute equal number of genes and produce similar effects.

But expression of some genes is significantly affected by the parental type, i.e. whether it came from male or female. This phenomenon is called Genomic Imprinting.

One of the example of genomic imprinting is Igf2 (Insulin like growth factor II). It is present in both mice and humans and in both cases genomic imprinting takes place.

A child inherit one allele from male parent and other allele from female parent. The paternal copy of Igf2 is actively expressed in the fetus and placenta but maternal copy remains completely silent.
So, when the paternal copy is deleted in the mice, the mice produces small placenta and low weight baby.

Another example would be Prader-Willi and Angelman syndromes. A child with Prader-Willi syndrome are mentally retarded, has small hands and feed and poor sexual development. Angelman syndrome has completely different symptoms like frequent laughter, uncontrolled muscle movement and a large mouth. Both the syndrome is caused due to deletion at the same region on long arm of chromosome 15. The only difference is if the chromosome with deletion is inherited from father, will result in Prader-Willi syndrome were as if the chromosome with deletion is inherited from mother, will result in Angelman syndrome.

The exact mechanism of genomic imprinting is still under investigation, but different methylation of DNA is one of the reasons behind it.

Genomic imprinting is one form of epigenetics, where process like acetylation of histone proteins, or phosphorylation and methylation of histone proteins or methylation of DNA itself result in a different phenotype. For example, even though homozygotic twins have same genotype, they differ in many ways because of epigenetics.

Friday, April 12, 2013

Hidden Markov Models


Hidden Markov Models
Hidden Markov Models is one of the popular machines learning approach that can be used to decipher highly complex biological data.
One example would be CG-Island.

CG is the least frequent nucleotide in many genomes


Reason:
C within CG is easily methylated
Resulting methylation has tendency to mutate into T
However methylation is suppressed around genes in areas called CG-Island. Here CG appears relatively frequently.

Fair Bet Casino:
Through it CG-Islands can be modeled.
Dealer flips coin and player bets on outcome: heads or tails
Dealer uses either a biased coin (head with probability ¾) or fair coin head or tail equally likely
Given a sequence of coin tosses, problem to find out when dealer used a biased coin or a fair coin
Obviously, if you observe long line of heads it’s likely that dealer used biased coin and if you observe even distribution of heads and tails he likely used fair one.