Professor Myles Lewis

Bioinformatics and machine learning in stratified medicine

Professor Lewis leads a bioinformatics and biostatistics team involved in analysing and integrating multi-omic data including bulk & single cell RNA-Sequencing (RNA-Seq), genotype & expression Quantitative Trait Loci (eQTL) analysis, immunoglobulin repertoire analysis, CyTOF and protein microarray. Our group has interests in machine learning to predict clinical response for personalised/ stratified medicine, data visualisation of big data and disseminating results through interactive data web portals. Our group developed a number of R/shiny apps (https://peac.hpc.qmul.ac.uk/) to allow exploration of RNA-sequencing data on synovial biopsies from the Pathobiology of Early Arthritis Cohort (PEAC). This website allows researchers to directly compare synovium and blood genes and gene modules in early rheumatoid arthritis and see how gene expression correlates with clinical measures of disease activity and response to therapy, thus enabling in-depth interrogation of the data. In the R4RA clinical trial, our analysis of RNA-Seq data from synovial biopsies from this cohort showed that patients with low levels of B cell signature genes were less likely to respond to the B cell depleting agent rituximab in comparison to the IL-6 receptor inhibitor tocilizumab. 

Ubiquitination in autoimmune disease

My lab research involves understanding the effects of ubiquitination pathways on the immune system, and their impact on autoimmune rheumatic diseases especially SLE (lupus), rheumatoid arthritis (RA) and other connective tissue disorders. My research focuses on trying to understand how genetic variation in ubiquitination genes influences B cell differentiation and autoantibody production through regulation of the master transcription factor of inflammation, NF-kB, thus increasing susceptibility to lupus and autoimmune disease. Our work has important translational medical implications, since we aim to determine whether particular ubiquitination genes represent novel targets for new therapies to treat lupus, RA and other autoimmune diseases.

Software

Software packages in R and python published by Professor Lewis’s team include:

R packages available from CRAN

• Spectrum performs fast self-tuning spectral clustering using an adaptive density aware spectral kernel.
• volcano3D enables visualisation of 3-way analyses of gene expression and other high dimensional 3-class data using 3D volcano plots or polar plots.
• glmmSeq enables analysis of longitudinal RNA-Seq and other omics data using mixed effect models which can either be negative binomial generalised linear mixed effects models (GLMM) which are suitable for RNA-Seq data or Gaussian linear mixed effects models (LMM). Analyses are parallelised for speed.
• nestedcv implements nested k*l-fold cross-validation for lasso and elastic-net regularised linear models via the ‘glmnet’ package and many other machine learning models via the ‘caret’ package.
• locuszoomr generates publication-ready regional gene locus plots similar to those produced by the web interface LocusZoom, but running locally in R. Genetic or genomic data with gene annotation tracks are plotted via R base graphics, 'ggplot2' or 'plotly', allowing flexibility and easy customisation including laying out multiple locus plots on the same page.
• easylabel enables interactive labelling of scatter plots, volcano plots and Manhattan plots using a ‘shiny’ and ‘plotly’ interface. Users can hover over points to see where specific points are located and click points on/off to easily label them. Labels can be dragged around the plot to place them optimally.

R packages in Bioconductor

• enhancedVolcano generates publication-ready volcano plots with enhanced colouring and labelling.
• M3C performs Monte Carlo reference-based consensus clustering of high dimensional omics data.

R packages on GitHub

• DEGGs tests for differential gene-gene correlations across different groups of samples in count data from high-throughput sequencing assays, leveraging gene-gene interaction graph network information.

Packages published in collaboration with other teams

• shinyExprPortal is a visualisation tool which enables deployment of configuration file-based 'shiny' apps with minimal programming for interactive exploration and analysis showcase of molecular expression data.
• MUSTANG – Multi Stain Graph Attention Multiple Instance Learning – a graph neural network pipeline for whole slide image (WSI) classification using multi-stain embeddings and self-attention graph pooling.
• bbmix models RNA-seq reads using a mixture of 3 beta-binomial distributions to generate posterior probabilities for genotyping bi-allelic single nucleotide polymorphisms.

Publications

• Lewis MJ, Spiliopoulou A, Goldmann K, Pitzalis C, McKeigue P, Barnes MR. nestedcv: an R package for fast implementation of nested cross-validation with embedded feature selection designed for transcriptomics and high dimensional data. Bioinform Adv 2023; 3(1): vbad048. PMID: 37113250
• Henkin R, Goldmann K, Lewis MJ, Barnes MR. shinyExprPortal: a configurable 'shiny' portal for sharing analysis of molecular expression data. Bioinformatics 2024; 40(4): btae172. PMID: 38552327
• Gallagher-Syed A, Rossi L, Rivellese F, Pitzalis C, Lewis MJ, Barnes M, Slabaugh G. MUSTANG: Multi-stain self-attention Graph Multiple Instance Learning pipeline for histopathology Whole Slide Images. British Machine Vision Conference 2023. arXiv:2309.10650
• Gallagher-Syed A, Khan A, Rivellese F, Pitzalis C, Lewis MJ, Slabaugh G, Barnes MR. Automated segmentation of rheumatoid arthritis immunohistochemistry stained synovial tissue. Front Med Tech 2023 arXiv:2309.07255
• Vigorito E, Barton A, Pitzalis C, Lewis MJ, Wallace C. BBmix: a Bayesian Beta-Binomial mixture model for accurate genotyping from RNA-sequencing. Bioinformatics 2023; btad393. PMID: 37338536
• Sciacca E, Alaimo S, Silluzio G, Ferro F, Latora V, Pitzalis C, Pulvirenti A, Lewis MJ. DEGGs: an R package with shiny app for the identification of differentially expressed gene-gene interactions in high-throughput sequencing data. Bioinformatics 2023; 39(4): btad192. PMID: 37084249
• John CR, Watson D, Barnes M, Pitzalis C, Lewis MJ. Spectrum: Fast density-aware spectral clustering for single and multi-omic data. Bioinformatics 2020; 36(4): 1159-66. PMID: 31501851
• John CR, Watson D, Russ R, Goldmann K, Ehrenstein M, Pitzalis C, Lewis MJ, Barnes M. M3C: Monte Carlo reference-based consensus clustering. Sci Rep 2020; 10(1): 1816. PMID: 32020004