g

p

therefore treated as an essential gene cluster. Several density

n algorithms have been developed for identifying essential genes

some genome-wise transposon statistic [Langridge, et al., 2009;

t al., 2012; Yang, et al., 2017]. Different models for essential

covery are based on different assumptions of the nature of

on insertion profile. TraDIS discovers essential genes through

g a density function based on the transposon insertion sites per

istic [Langridge, et al., 2009]. ESSENTIALS discovers essential

rough estimating a density function based on the transposon

s per gene statistic [Zomer, et al., 2012]. DEM (distal effect

iscovers essential genes through estimating a density function

the mutation feature statistic which is the convolution between

poson insertions per gene and the transposon insertion sites per

ng, et al., 2017].

ddition to density estimation, cluster analysis can also be

ed when dealing with the problems for separating essential genes

-essential genes, especially in the context of multivariate analysis.

ity estimation

en, the future inference is required based on the knowledge

rom collected experimental data through a pattern discovery and

process. Various density estimation approaches are such a process

nstructing an unknown data distribution from which data are

to be sampled or drawn [Silverman, 1986; Duda, et al., 2000].

nce, a mean value and a standard deviation value can be estimated

a set if the data set is assumed to be a sample drawn from a

distribution. After these two parameters have been well-

d, a distribution model of the data set can be constructed and can

or the future inference for novel data.

uld be noted that an estimated density based on a collected data

have some deviation from the expectation. This is not a surprise

a drawn sample always has a much smaller size than a whole