Dr Mizanur R Khondoker PhD
|contact this person|
|tel||020 7848 0305|
|fax||0207 848 0281|
Institute of Psychiatry
NIHR BRC for Mental Health
Ph.D in Statistical Bioinformatics jointly from Biomathematics and Statistics Scotland (BioSS) and the University of Edinburgh.
M.Sc. in Statistics with specialization in Biostatistics from the University of Dhaka.
B.Sc. (Honors) in Statistics from the University of Dhaka.
Currently working as a Lecturer in Biostatistics at the NIHR Biomedical Research Centre for Mental Health at the South London and Maudsley NHS Foundation Trust and Institute of Psychiatry, Kings College London.
Previously served as a Statistics and Statistical Bioinformatics Research Fellow at the Division of Pathway Medicine, University of Edinburgh Medical School.
Also served as an Assistant Professor at the Department of Statistics, Biostatistics and Informatics, and as a Lecturer at the Institute of Statistical Research and Training (ISRT), University of Dhaka.
Teaching at the Institute’s general teaching programmes on Behavioural Research Methodology. The department provides a modular programme of courses on key statistical methods in behavioural research, training in statistical software and a contribution to meeting training requirements in introductory statistics.
Developer and maintainer of R packages
The R packages listed in this page are under GNU General Public License (version 2 or later) and freely available for academic use. The multiscan package is available in Bioconductor. The optBiomarker package is available from the Comprehensive R Archive Network (CRAN).
It is a common practice to scan a microarray several times at different scanner’s settings, but then to use data from a single scan chosen on the basis some informal judgements. We show in
Statistical estimation of gene expression using multiple laser scans of microarrays (Bioinformatics, (2006), that it is possible to reduce the sampling variability in gene expression estimates if they are estimated from multiple scans using appropriate statistical models. As scanning a microarray more than once does not require any extra resources, estimating gene expressions from multiple scans is a cost free gain in precision. The multiscan package implements the method we proposed in the above paper. This package can be downloaded from from Bioconductor repository.
optBiomarker: R package for estimating optimal number of biomarkers at a given error tolerance level for various classification rules
The package provides functions for simulating microarray data, computing generalisation errors for two-group classification problems based on four commonly used machine learning methods, namely, the Random Forest, Support Vector Machine, Linear Discriminant Analysis and k-Nearest Neighbour. Data can be simulated with different number of biomarkers and spot replications, and various levels of several other data characteristics, such as training set size, biological variation , experimental (technical) variation , differential expression (fold-change) and correlation between biomarkers.
The package also provides a database of generalisation errors, and a userfriendly R GUI with 3D real-time rendering system and control panel for interrogating the database with an aim to estimate the optimal number of biomarkers for given levels of predictive accuracy as a function of the above data characteristics.
The optBiomarker package can be downloaded from the Comprehensive
R Archive Network (CRAN).
to be updated
last updated: Friday, March 04, 2011