Blog posts

Next session – 27/03 – Multilevel Monte-Carlo

Next session: Wednesday 27th March, with an exciting paper on a topic new to us!

Paper: “Multilevel Monte Carlo methods”, M. B. Giles, Acta Numerica, 2015, vol. 24, pp. 259-328

https://www.cambridge.org/core/journals/acta-numerica/article/multilevel-monte-carlo-methods/C5AF9A57ED8FF8FDF08074C1071C5511

Monte Carlo (MC) methods are a very general and useful approach for the estimation of expectations arising from stochastic simulation. However, they can be computationally **very** expensive. Multilevel Monte Carlo is a recently developed approach which greatly reduces the computational cost by performing most simulations with low accuracy at a correspondingly low cost, with relatively few simulations being performed at high accuracy and a high cost. This article reviews the ideas behind multilevel MC, and various cool applications.

Very relevant for infectious disease modeling!

Presenter: Juliette Unwin

Wed. 04/03 – PGS and LASSO paper

please join us this coming Wednesday 6th March for the next round of the Stats reading group!

Where: MSc student room

When: 11am

Presenter: Ville Karhunen

We will be looking at the following paper

“Polygenic scores via penalized regression on summary statistics”, Mak et al, 2017. Genetic Epidemiology,  vol. 41 issue 6

Polygenic scores (PGS) summarize the genetic contribution of a person’s genotype to a disease or phenotype. They can be used to group participants into different risk categories for diseases, and are also used as covariates in epidemiological analyses. Recently, there is much interest in methods that use published summary statistics. However there is no inherent information on linkage disequilibrium (LD) in summary statistics, so we have to use LD information available elsewhere. The authors propose a method for constructing PGS using summary statistics and a reference panel in a penalized regression framework.

https://doi.org/10.1002/gepi.22050

Journal club + Seminar

Please join us tomorrow 30/01 for the next round of the Stats reading group at SMH, with journal club + seminar sessions !

All in the MSc students room, ground floor of SMSM building

11am-midday: Paper appraisal by students from the MSc in HDA&ML: “Discovering regulatory and signalling circuits in molecular interaction networks”, T. Ideker et al, BioInformatics, 2002.

https://pdfs.semanticscholar.org/5f90/92575e1bcafa69ab360334570166d3d8b16b.pdf

In this paper, the authors propose a model for the integration of univariate associations results with protein-protein and protein-DNA interactions stored in large and public databases. The method is based on (i) the construction of a network model using a priori knowledge on molecular interactions and (ii) the estimation of univariate z-scores representing the association of each gene with the outcome. The z-scores are then aggregated for each subnetwork in order to identify the most relevant subnetworks in association with an outcome of interest.

midday-1pm: Seminar by Benno Schwikowski, author of this week’s paper

Benno Schwikowski is the head of the Systems Biology group at Institut Pasteur (France). He has worked a lot with network models and in particular he developed Cytoscape (a software for network visualisation). In his seminar, he will present LEAN, a method for identification of subnetworks enriched in nodes associated with an outcome of interest.

New year new sessions!

Happy new year 2019!

And welcome back! We are kick-starting this new year and the 2nd term with fresh new round of the SMH stats reading group:

Venue: MSc student room, ground floor of SMSM building

Time: 11am

2nd meeting – 07/11

Paper of the week:

“metaCCA: summary statistics-based multivariate meta-analysis of genome-wide association studies using canonical correlation analysis”, Cichonska et al. ,2016. Bioinformatics, Volume 32, Issue 13, Pages 1981–1989

https://doi.org/10.1093/bioinformatics/btw052

Presented by: Dr Schneider-Luftman

GWAS (Genome-wide association study) is a method widely used nowadays for the analysis of association between the genome and any phenotypic trait. In its classical form, it works as a “giant” univariate regression of 1 outcome against the allelic variation of all the SNPs (Single-nucleotide polymorphism) in the genome *. (*= except rare SNPs, chrm. X and Y, and other conditions….). However, there is increased interest in analysis multiple related traits together. It only makes sense to analyse them jointly, in order to increase statistical power.

This paper introduce a framework to perform multi-outcome meta-analysis of GWA data, from a single of several studies. The underlying method relies on CCA (canonical correlation analysis) + shrinkage. The paper also discusses an application to several Finnish population cohorts – notably the Northern Finland Birth Cohort, very widely used at EBS – across 81 lipid NMR (nuclear magnetic resonance) measures.

1st meeting of 2018/19 tomorrow!

Wednesday 24/10 @ 11am, in MSc students room (ground floor)

Do come along to the 1st Stats reading group meeting of the year, and learn about matrix completion & compressed sensing:

Exact Matrix Completion via Convex Optimization , E. J. Candes & B. Recht, 2008

Presented by Dr Samir Bhatt

Suppose we have a (largely unknown)  nxn matrix but we can only see m << nxn entries. Can we retrieve the whole matrix? Yes, yes we can (under some specific conditions)!

Paper 1 announced!

On the first session of the Stats reading group, we will be reviewing the following:

Exact Matrix Completion via Convex Optimization (E. J. Candes & B. Recht, 2008)

Presented by: Dr Samir Bhatt

“Matrix completion is the task of filling in the missing entries of a partially observed matrix. Sounds like imputation right? That’s useful as we always have missing data (damn)! In this group meeting, We will discuss one of the two seminal papers that have sparked a new field of interest called compressed sensing”

A cool paper on compressed sensing X matrix algebra

Disclaimer – In this reading group, we are about learning new things, and none of us are experts in the field of compressed sensing. Attendants are strongly encouraged  not to be intimidated and to attend, even if the materials feels very unfamiliar