Computational Systems Biology Methods and Protocols.7z

3 Methods

3.1 Variants of
Machine Learning
Methods in Biological
and Biomedical Study

To apply machine learning methods in biological study, different

variants have been proposed to address particular problem appeared

in the biological questions. According to the biological back-

ground, there will be several application scenarios, e.g., sequence

analysis, image analysis, interaction analysis, disease analysis, and

annotation analysis. The following introduction will illustrate the

wide strategy of applying machine learning in each scenario, rather

than conventional survey on models of machine learning.

3.1.1 Application in
Sequence-Focused
Analysis

Based on general-purpose machine learning algorithms and

libraries, many software packages have been designed to learn

genotype-to-phenotype predictive models from sequences with

known phenotypes [46, 47], which can computationally judge the

genetic bases of phenotypes [36]. One is to recognize the regu-

latory elements on the biological sequence: to select appropriate

features of promoters that distinguish them from non-promoters,

the nonlinear time series descriptors along with nonlinear machine

learning algorithms, such as support vector machine (SVM), are

used to discriminate between promoter and non-promoter regions

[48]; and a machine learning approach, MutPred Splice, has been

developed to recognize coding region substitutions that disrupt

pre-mRNA splicing, which can be applied to detect the splice site

loss [49]; and a classifier is built and trained by using the enhancer

set and identified related enhancers based on the presence or

absence of known and putative TF binding sites, which combine

the machine learning and evolutionary sequence analysis [50]; and

an ortholog prediction meta-tool, WORMHOLE, is to integrate

distinct ortholog prediction algorithms into meta-tools to identify

novel least diverged orthologs (LDOs) with high confidence

[51]. Two is to predict important proteins from sequences due to

the difficulty on wet experiment: the top-performing methods

based on machine learning approaches have been built to tackle

both the detection of transmembrane beta-barrels in sets of

Table 2
(continued)

Methods Description URL

Metaml The software framework can analyze microbiome profiles

and metadata for thousands of samples [44]

http://segatalab.cibio.

unitn.it/tools/metaml

Hierarchical

boosting

A machine learning classification framework can combine the

selection tests to detect the features of polymorphism in

hard sweeping with controls on population-specific

demography [45]

http://hsb.upf.edu/

Revisit of Machine Learning Supported Biological and Biomedical Studies 189