Systems Biology and Learning Machine

This is a short paper about Systems Biology and Machine Learning.
As a matter of fact, it is pointed out herein the natural synergy between them. We use as default reference and leave further references throughout the document.
Introduction
See also: Systems biology, Machine learning
Systems biology strives to tackle with complexity by studying how "simple" systems interact, what gives rise to the well-known emergent properties. Emergent properties is an outcome of the new paradigms that overlook the reductionism, see for example for discussions on that direction.
The viewpoint that systems biology prizes is the one named "holistic", discussed for a long time in the physics community, see. This creates a necessity to handle situation not so common to "classical mathematics", such as genetic networks, protein networks, and so on.
On the other hand, the necessity to handle situation uncommon to classical mathematics has been present since last decades. See for example the simple problem of pattern classification; even for simple problem such as the problem of separating two groups might be difficult. Hence, in layman terms, machine learning in systems biology attempts to exploit several tools that are model-independent such as neural networks, or even bayesian networks.
Works on the literature
The concept of machine learning in systems biology is not new at all, see for example programmes in systems biology such as master level or even PhD. See for example the references,.
Maybe was is really provokative is the proposal of computational intelligence as an embedded methodology to systems biology. This is justified mainly based upon the kind of problems faced on the area, in general just mathematics does not suffice. See for several examples of application of computational intelligence.
As a complement for the discussions in the previous paragraph, studied the problem of motif discoveries, structures that repeats again and again, in unaligned DNA and protein sequence . They proposed a self-organizing neural network structure for solving the problem of motif identification in DNA and protein sequences. The aim of their work is to develop an algorithm that can identify and locate motifs, if any, in a DNA or protein sequence. They use the idea that the “alphabet” of the DNA molecules is made up of four letters while for protein it is 20 letters. They define a motif as a short stretch of a molecule that forms a highly constrained sequence. applied ANN to locate the promoter in DNA. They made the comparison with well-known techniques, achieving good results. In the article, the authors compare an inductive learn algorithm (called Rule 3) with ANN.
Machine learning and microArrays
See in addition Microarray, see for more discussions.
Loosely stating, microArray is a matrix of colors, where the color is correlated with the gene expression, see for example. Therefore, we can apply supervised learning for training pattern identification, for example, or use unsupervised learning for finding patterns that could correspond to a network motif or even a cromossome.
See also
* Systems Biology
* Transcription network
* Artificial Neural Networks
*
*

Main Menu

See Also

Main Menu