Code to reproduce results of paper "Modeling integration site data for safety assessment with MELISSA"
- analyses code and data for analyses,
- figures where figures are stored,
- package implementation of MELISSA,
- plots code to produce figures paper,
- simulation code and data for simulation.
It is necessary to set the path of this folder in all files in which there are operations to read and store data. These files are:
- all files in the folder plots (6 files),
- all files in the folder analyses/code (18 files),
- files simulation/control.R and simulation/code/computePpvCov.R.
For all these files, the path of the current folder is stored in the first line of code:
mainFolder = "~/Downloads/MELISSApaper"
with the object mainFolder that is then used to construct paths to read and store data in all mentioned R files. Change the path to run these files.
The R packages required are data.table (for most files); stats, fastglm, statmod, parallel, glm.fit, brglm.fit (to run analyses); ggplot2 (for all plots); annotatr, IRanges, GenomicRanges, vegan, enrichplot, clusterProfiler, ReactomePA, biomaRT, pheatmap, rtracklayer, nVenn, readr, ggtext, ggstance, purrr (required for one or more plots).
Run the files fig2.R, fig3.R, fig4.R, fig5.R, fig6.R, fig7.R in the plots folder, the output of these files are saved in the sub-folders of figures.
The file fig2.R uses the output of a simulation design that are stored in the sub-folders of smiulation/resu. The files fig3.R, ..., fig7.R use data that are stored in the sub-folders of analyses/data and analyses/results.
The epigenetic data required for Figure 6D are publicly available but they are not provided here. The file analyses/data/epig/downloadForFig6D.txt contains the information on which datasets should be downloaded and stored in analyses/data/epig/ to reproduce Figure 6D.
Run the file simulation/control.R after setting the values for:
- type (in line 3) which can be 1, 2, 3, 4 for the analysis of gene targeting, gene differential targeting, clone fitness, clone differential fitness respectively.
- size (in line 4) which can be 1, 2, 3, 4 to set the size of the samples, these values correspond to the sizes 100, 200, 400, 800 respectively for the targeting analyses, and 1000, 2000, 4000, 8000 respectively for the fitness analyses.
- heig (in line 5) which can be 1, 2, 3, 4, 5, 6, 7 to set the size of the effect, these values correspond to the effect sizes 1 (no effect), 2, 4, 8, 16, 32 and 64 respectively.
- nSimulations (in line 10) which sets the number of simulations. In the paper this was equal to 1000.
The output of the simulations are stored in the sub-folders of simulation/simu. For the analysis specified by type, after running the simulations for all values of size and heig, the function simulation/code/computePpvCov.R can be used to compute and overwrite the files in the sub-folders of simulation/resu which are used to compute the plots in Figure 2.
Run the .R files inside the folder analyses/code. Running these files will replace the omonimus .csv objects in analyses/results.
The analyses used in Figures 4 - 7 are the following:
- 4: gt_ov_hspc, gt_ov_bmsc, gt_ov_amsc (gene target analysis, for HSPC, Bone marrow MSC, Adipose MSC, data respectively).
- 5: gt_di_bmsc_vs_hspc (gene differential target analysis, group 1: BM MSC, vs group 2: HSPC), gt_di_amsc_vs_hspc (Ad MSC vs HSPC), gt_di_bmsc_vs_amsc (BM MSC vs Ad MSC).
- 6: cf_ov_bmsc, cf_ov_amsc (clone fitness analysis, for BM MSC, and for Ad MSC, respectively), cf_di_bmsc_vs_amsc (clone differential fitness, group 1 BM MSC, group 2 Ad MSC).
- 7: gt_ov_b0be_lym, gt_ov_bsbs_lym, gt_ov_was__lym (gene target analysis, for lymphoid (B, T, NK) cells of B-THAL, SCD, and WAS patients, respectively), gt_ov_b0be_mye, gt_ov_bsbs_mye, gt_ov_was__mye, (gene target analysis, for myeloid (Granulocytes, Monocytes) cells, of same patients), gt_ov_hspc_all (for HSPC), cf_ov_bsbs_all (clone fitnes for all cells (B, T, NK, Granu, Mono) of SCD patient), gt_di_2_was__vs_hspc (gene differential target, group 1 all cells of WAS patients, group 2 HSPC).
The datasets used are in the folders analyses/data/expr and analyses/data/pati for experimental and clinical data, respectively.
Title: Modeling integration site data for safety assessment with MELISSA
Authors: Tsai-Yu Lin, Giacomo Ceoldo, Kimberley House, Matthew Welty, Thao Thi Dang, Denise Klatt, Christian Brendel, Michael P. Murphy, Kenneth Cornetta, Danilo Pellin.
Abstract: Gene and cell therapies pose safety concerns due to potential insertional mutagenesis by viral vectors. We introduce MELISSA, a statistical framework for analyzing Integration Site (IS) data to assess insertional mutagenesis risk. MELISSA employs regression models to estimate and compare gene-specific integration rates and their impact on clone fitness. We tested MELISSA under three settings. First, we conducted extensive simulation studies to verify its performance under controlled conditions. Second, we characterized the IS profile of a lentiviral vector on Mesenchymal Stem Cells (MSCs) and compared it with that of Hematopoietic Stem and Progenitor Cells (HSPCs), in addition to comparing the in vitro clonal dynamics of MSCs isolated from alternative tissues. Finally, we applied MELISSA to published IS data from patients enrolled in gene therapy clinical trials, successfully identifying both known and novel genes that drive changes in clone growth through vector integration. MELISSA identifies over- and under-targeted genes, estimates IS impact, analyzes differential targeting, and explores biological relevance through pathway analysis. This offers a quantitative tool for researchers, clinicians, and regulators to bridge the gap between IS data and safety and efficacy evaluation, facilitating the generation of comprehensive data packages supporting Investigational New Drug (IND) and Biologics License (BLA) applications, and the development of safe and effective gene and cell therapies.