Background and history
Profiling Relative Inhibition Simultaneously in Mixtures (PRISM) is a unique, novel technology that enables the rapid screening of thousands of drugs against over 900 human cancer cell line models using a high-throughput, multiplexed approach.
PRISM was first conceptualized by researchers at the Broad Institute of MIT and Harvard in 2012 and scaled into a high-throughput screening platform in 2015. Since then, the PRISM team has screened over 30,000 drugs with over 500 cell lines, and worked with 200+ collaborators from academia, biotech, and big pharma.
First publication of the PRISM method from 2016. 102 cell lines were labeled with unique 24-nucleotide barcodes and used to screen 8400 compounds.
An updated panel of 578 barcoded cancer cell lines were used to profile the activity of 4518 existing drugs. Results were used to begin the PRISM drug repurposing resource.
The PRISM assay workflow
Cell line barcoding and pool preparation
The PRISM assay uses over 900 human cancer cell lines with deep multi-omic characterization.
Cell lines were stably transfected with unique DNA barcodes and quality control tested prior to use in the PRISM assay to:
- Verify the absence of mycoplasma
- Verify cell line identity with SNP fingerprinting
- Confirm unique 24-nucleotide cell line barcode identities
Cell lines are mixed into assay-ready pools of 20-25 cell lines based on doubling time similarity. Pools of cell lines are then frozen so they can be easily thawed for screening.
5-day viability assay
Test agents are added to 384-well plates at 8 doses with 3-fold dilutions and screened in triplicate against pooled cell lines for 5 days. At the end of incubation, cells are lysed, and mRNA is isolated. Barcode sequences are amplified by PCR and quantified using a Luminex FlexMap 3D® scanner. Relative barcode abundance (compared to vehicle control) is then used to generate cell line sensitivity profiles for each test agent and concentration tested.
Data analysis and predictive modeling
PRISM sensitivity profiles are compared to deep multi-omic cell line characterization information in univariate analyses as well as multivariate predictive modeling algorithms to identify features that correlate with sensitivity.
The Broad Institute feature sets used to understand the differences between sensitive and insensitive cell lines include:
- Cell line lineage and subtype, mutation status, gene expression, protein expression, copy number, methylation, and metabolomics (Cancer Cell Line Encyclopedia)
- Loss-of-function genetic perturbation data using CRISPR and shRNA (Cancer Dependency Map and Project Achilles)
- PRISM viability data from over 6,000 annotated commercial drugs (Drug Repurposing Hub)
A current list of datasets used for each of the feature analyses can be found on our GitHub repository, and are available for download in the DepMap portal.
Drug combination screening
Combination screening requires careful selection of drug doses which can be especially difficult in a pooled context. Therefore, it is only recommended to use this assay for test agents with previous validation and/or known dosing. When selecting an anchor dose our recommendation is to select a dose that does not broadly affect cell viability but gives a reproducible phenotypic effect in a specific cell line or set of cell lines.
All combination screens are run in the following format:
- Drug A at dose (7 pt. dose with 3-fold dilutions) alone
- Drug B at one anchor dose alone
- Drug A at dose (7 pt. dose with 3-fold dilutions) plus Drug B at one anchor dose
This screen of Drugs A and B counts as two ‘slots’ in the PRISM assay.
In addition to the standard DMSO-based screening format, PRISM now has the ability to screen aqueous small molecules and antibodies.
This new offering is enabled by PRISM’s commitment to research and development and to continuous improvement of assay technologies and expansion of screening capabilities.
Cell line collection
As a cell line screening assay, PRISM continually strives to improve our cell line collection to ultimately have the largest, most robust, and most diverse cell set possible. By partnering with academic groups and cell banks across the globe, we hope to capture and reflect the diversity of human cancers. PRISM’s PR1000 cell collection contains over 900 barcoded cell lines representing more than 45 lineages.
See a full annotated list of our PRISM barcoded cell lines.
Cell sets and pools
The full collection of PRISM cell lines is screened for all standard PRISM experiments. This PR1000 collection is divided into two cell sets:
- PR500: 488 solid tumor adherent cell lines
- PR300+: 431 hematopoietic, solid tumor, and pediatric cell lines
Within each cell set, cell lines are pooled in groups of 20-25 based on their growth rates. For the standard PRISM assay, each well contains a single pool of 20-25 cell lines.
See the full cell set here.
Cell culture conditions
In order to grow together in pooled conditions, all cell lines have been adapted to a single, shared media: RPMI-1640 without phenol red (catalog link here). No supplements or antibiotics are added other than fetal bovine serum (FBS) which is added at 10% for adherent cell lines and at 20% for suspension cell lines.
All standard PRISM assay 384-well plates are prepared by seeding adherent pools at 1,250 cells per well and suspension pools at 2,000 cells per well.
The Broad Institute has made PRISM available to all collaborators.
We are excited to collaborate with you to advance your research to find new treatments for cancer patients. Together we can provide a better understanding of cancer therapeutics and accelerate the drug discovery process.
PRISM's high-throughput drug sensitivity assay
Traditional phenotypic screening is done one cell line at a time, requiring lots of time and reagents, which can slow down drug discovery and development efforts. PRISM’s unique approach to barcoding and pooling cell lines for high-throughput screening overcomes the limitations of single cell line studies, helping researchers rapidly collect cancer cell line drug-sensitivity data on an unprecedented scale.
PRISM screening can be used to:
- Uncover novel drug mechanisms of action
- Identify unexpected off-target effects and toxicities
- Establish which patient populations might benefit from treatments
Industrial collaborative working model
PRISM’s mission is to help improve cancer patients’ lives. To that end, we are excited to collaborate with industrial partners to advance their understanding of their cancer therapeutics. Our working model is that of a scientific collaboration to further cancer research.
Some key points of our collaboration agreement include:
- Data Ownership – PRISM retains the right to use collaboratively generated data to validate and improve the PRISM platform
- Intellectual Property – adherence to U.S. patent law
- Publication – data may be jointly published when both partners agree
PRISM invites collaborators to submit test agents to three consortium-style screens each year. These screens use the standard PRISM study design (PR1000 cell set, 5-day assay, test agents run at 8 pt. dose with 3-fold dilutions in triplicate). Types of submissions accepted include small molecules, antibodies, and combination studies. Data is returned in approximately 4 months.
To participate in one of these screens, collaborators must:
- Fill out the submission form during the submission window
- Ship a stock of 150µL of each test agent at 1000x the top screening concentration
- Provide a purchase order to cover the cost of screening
- Sign the PRISM collaboration agreement (industry partners only)
Sign up for our mailing list to receive notifications about upcoming submission windows.
Barcode abundance is compared to vehicle control conditions to generate relative cell line sensitivity signatures.
These sensitivity profiles are then correlated to baseline feature sets using univariate and multivariate analyses to identify features that are associated with drug response.
PRISM quality control
PRISM provides QC data for every cell line used in the assay. PRISM quantifies the separation of results obtained from the negative control (vehicle only) and positive control (a high-dose of the pan-cytotoxic drug bortezomib for DMSO screens or puromycin for aqueous screens) using error rate and dynamic range. Cell lines that do not meet performance standards are removed from all downstream analyses.
PRISM viability assay
PRISM includes validation agents in each screen with known mechanisms of action (MOA). These well-characterized, commercial reagents produce robust, reproducible sensitivity profiles and feature correlations which are used to ensure high cross-screen reproducibility.
Here we provide an example of PRISM data for two frequently screened compounds:
- AZ628, a BRAF inhibitor
- Imatinib, an inhibitor of PDGFRA and the BCR-ABL1 fusion protein
Melanoma cell lines containing the hotspot BRAF V600E mutation are more sensitive to AZ628 than BRAF wild-type cell lines. CML cell lines, which contain BCR-ABL1 fusions, and lines that have high expression of PDGFRA are selectively killed by imatinib, as expected.
PRISM predictive modeling
Predictive models are run on available feature sets using
- CCLE genomic data
- Gene expression data
- Mutation and copy number data
- Loss-of-function genetic perturbation data from the Dependency Map
- Viability data from the PRISM drug repurposing project
PRISM identifies the expected biomarkers for validation agents. The example below shows the identification of BRAF as a top dependency from both the RNAi and CRISPR datasets, along with other RAF inhibitors from the publicly available PRISM Repurposing sets. Similarly for Imatinib, PDGFRA, BCR, and ABL score as the top correlated genetic dependencies along with other PDGFR inhibitors.
We provide a detailed report for the quality of the assay, viability of all 900 cell lines, feature set correlation analysis, and validation compound reports as well as all the raw data from your test agents. See below for more details.
Standard data release includes
- Summary of analyzed data for each test agent
- Log-fold change
- Viability of each cell line at each drug dose (displayed as a heatmap)
- Individual dose data
- Computed for each cell line that displays sensitivity to drug treatment
- Used to calculate parameters like AUC and IC50
- Raw and normalized fluorescence values
- Viability data (and replicate-collapsed viability data)
- Detailed cell line quality control data for all test agents
Feature set correlations
- Interactive plots and tables for univariate (continuous and discrete associations) and multivariate (Random Forest model) analyses▪ Lineage, mutation, gene expression, CRISPR knockout, repurposing, etc.
- Data correctly formatted for upload into the DepMap data explorer tool
Combination screen assay deliverables
Three reports are delivered for each combination screened:
- Analysis of the test agent at dose and anchor separately
- Analysis of the combination
- Analysis of the differential response between single agent and combination
Single-agent DMSO screens
- 3 validation compounds run in parallel:
- AZ-628, a pan-RAF inhibitor
- Imatinib, a BCR-ABL1 kinase, KIT, and PDGFR receptor tyrosine kinase inhibitor
- Nutlin-3a, an inhibitor of the MDM2-p53 interaction
- 1 known synergistic combination run in parallel:
- Temozolomide, a DNA alkylating agent + O6-benzylguanine, an alkyltransferase inhibitor
- 2 validation compounds run in parallel:
- Imatinib, a BCR-ABL1 kinase, KIT, and PDGFR receptor tyrosine kinase inhibitor
- Palbociclib, a CDK4/6 inhibitor
Public PRISM reference data sets
In collaboration with the Broad’s Dependency Map team, public PRISM data has been generated for thousands of commercially available drugs. These repurposing screens are accessible on the DepMap portal, which features 6000 drugs with diverse mechanisms of action and indications.
PRISM is committed to further develop and improve our cutting-edge screening technologies. To update our assay for new screening modalities, we are constantly looking at the market to find out what kinds of new therapies companies are developing. We then try to develop high-throughput pooled screens to see if we can screen these new modalities in PRISM. Some caveats that we need to consider are paracrine effects between cell lines in the pools, faster cell lines overgrowing the pools in longer-term assays, and our high-throughput workflow that requires multiple freeze thaws of the test agents.
As we develop the ability to screen new kinds of therapeutics, we will create public datasets for the research community to better understand how they work across hundreds of cell lines.
Another area of research for PRISM is technology development. We want to stay on the cutting edge of cancer research. A question that we are always asking is
“How can PRISM help expand our understanding of cancer?”
One area that many groups are working on now, and that we hope to bring another dimension to, is modeling the tumor microenvironment. We know that two-dimensional cell lines are very limited in what answers they can provide and therefore are working hard on new assays to better mimic the tumor microenvironment. We are starting to work on 3D spheroid and co-culture assays and are looking for collaborations in these areas.
Read more about PRISM in our recent publications.
Stay informed about new developments at PRISM.