Andrea Bryan, Timothy Erps, Arturo Escajeda, Anthony Fusco

High-Sensitivity Multiplexed Assay and Associated Bioinformatics Tools for Measuring
and Analyzing up to 68 microRNAs from Serum, Plasma, and Urine in 96-Well Format
Andrea Bryan, Timothy Erps, Arturo Escajeda, Anthony Fusco, Jeffrey Horne, Arben Koshi, Davide Marini,
Daniel Pregibon, Conor Rafferty, Isaac Stoner, Mike Tackett, Jessica Tytell, Andreas Windemuth
Firefly BioWorks, One Kendall Square, 1400W, Cambridge, MA 02139
Abstract
We used reference RNA from three tissue types in order to benchmark the Firefly™
microRNA HS assay with existing profiling methodologies, showing that assay correlates
well (Pearson > 0.8) with each. Prior to Pearson calculations, samples were first mean
normalized (signal for a given target in a sample was divided by the average signal of all
targets in that sample). Once target was normalized, signals were sample normalized by
dividing the target signal in a given sample by the average signal across samples for that
particular target. The Firefly™ HS microRNA assay was also compared to results obtained
on the NanoString (NS) system and gave a Pearson correlation of 0.8.
We isolated RNA from the serum samples (200ul) of three healthy individuals (A -C) using
a modified Trizol-LS protocol (a negative control isolation from water was also
performed). The Firefly™ miRNA HS assay was performed to simultaneously quantify 30
miRNAs in each sample with final assay readout on a Guava easyCyte 8HT. The results
from duplicate isolations were averaged and mean normalized. The miRNA profiles
show a broad range of miRNA levels with robust detection of most targets assayed.
Arbitrary Units
100
A
B
10
C
1
TLDA
Sequencing
Micro Array
Firefly HS
Assay Overview
TLDA
1 (0)
0.85 (0.02)
0.83 (0.1)
0.79 (0.14)
Sequencing
Micro Array
1 (0)
0.81 (0.1)
0.81 (0.08)
1 (0)
0.84 (0.1)
Integrated Bioinformatics
Firefly HS
1 (0)
Reproducibility
We performed the assay on total RNA purified from 200ul of human serum by a modified
Trizol protocol in quadruplicate in order to assess reproducibility. Targets present at
abundances at the middle or top end of the dynamic range (> 10 zeptomoles) give
coefficients of variation typically less than 10% (Figure A). Variation increases as signals
approach the Limit of Detection but typically remain < 50% (Figure B).
A
High Sensitivity Technical Replicates (High Abundance Targets)
Target and Sample
Normalized Signal (AU)
Hydrogel particles made using Firefly BioWorks' patented Optical Liquid Stamping are
hydrophilic, bio-inert, and flexible. The particles (shown in A) are porous, allowing target
molecules to diffuse and bind in a unique nanoscale three-dimensional scaffold. This
both maximizes signal and widens the assay's dynamic range. The particles are
“barcoded” with different levels of fluorescence in each of tow barcode regions (shown
in B), forming up to 70 unique codes. Each uniquely-barcoded particle is then
functionalized with an individual probe complementary to the microRNA of interest
enabling multiplexing of up to 68 custom targets along with a positive and negative
control in each well of a 96-well plate, all with a short 4-hour workflow (shown in C). The
Firefly™ HS microRNA Assay uses a unique post-hybridization labeling scheme followed
by bias-free PCR with universal primers to selectively amplify only captured targets.
Results are read out by flow cytometry and analyzed using our free
bioinformatics-enabled software tools.
microRNA Profiling in Serum
1000
5
4
3
2
1
0
B
Target and Sample
Normalized Signal (AU)
Firefly BioWorks has created a miRNA profiling assay that
enables the profiling of up to 96 samples in one afternoon
with PCR-level sensitivity. The assay is completely
customizable, allowing the detection of up to 68
user-selected miRNAs per sample. The platform is built on
barcoded microgels that can be scanned on existing flow
cytometers. We demonstrate how this assay can be used for
the highly sensitive detection of miRNAs in tissue, plasma
and other sample types, generating robust miRNA profiles
with as little as 1ng input.
We also introduce
bioinformatics tools that are integrated with the assay
analysis software to cluster co-regulated microRNAs and
help identify potentially relevant genes and pathways as well
as recommend additional microRNAs for study.
Cross-Platform Comparison
High Sensitivity Technical Replicates (Lower Abundance
Targets)
Biomarker discovery and validation typically involves additional challenges in the area of
target selection and data analysis. Firefly has created two bioinformatics tools to aid both
selection and analysis of microRNA targets: the Firefly™ Discovery Engine and the
Firefly™ Analysis Workbench. The web-based Firefly™ Discovery Engine automatically
generates a list of the most relevant microRNAs and associated gene targets for any
keyword or topic, returning a list of the most published microRNAs for that topic with an
annotated list of associated publications. The Firefly™ Analysis Workbench analyzes
particle scan data providing means for rapid data analysis and visualization. The
Workbench enables streamlined generation of publication-ready heatmaps and
bargraphs. Clusters of miRNAs identified in a study are cross-referenced with genomic
databases to provide further insghts into associated genes and pathways. Together, this
novel combination of bioinformatics tools and multiplexed high-sensitivity assays
enables rapid validation and screening of microRNA biomarker signatures from serum,
plasma, and urine in high throughput format.
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
Firefly™ Discovery Engine
www.fireflybio.com/search
Robustness and Input Amount
We demonstrate robust miRNA profiling over a broad range of input RNA from 100
down to 1ng. Without normalization (A), signals decrease linearly with decreasing
sample input. After Quantile Normalization is applied (B), the profiles for all sample
input leves are virtually identical. As expected, the correlation between the samples
are very strong, with Pearson coefficients >0.95. These results demonstrate the sensitivity and dynamic range of the assay, but also show that the assay is not biased by
input amount.
Sensitivity Analysis
Firefly™ Analysis
Workbench
A. Non-normalized
C. Correlation between
input amounts
We assessed the sensitivity of the assay using synthetic microRNA spike-ins across a
range from 1 attomole (600,000 molecules) to 1 zeptomole (600 molecules). Particles
containing a negative control probe, cel-miR-39-3p, were also included in all
experimental wells. All target levels assesed yielded detectable signal with excellent
linearity across the range tested. Error bars are shown between technical replicates.
100 ng
10 ng
1 ng
100 ng
1
0.9894
0.9574
10 ng
1 ng
1
0.9646
1
100 ng vs 10 ng
hsa-mir-141-3p
10000
10000
B. Normalized
hsa-mir-143-3p
1000
hsa-mir-203
cel-let-7-5p
100
10
100
Arbitrary Units
hsa-mir-210
R² = 0.9894
1000
10
1
0.1
1
0.1
10
100
1000
10000
Integrated
Cluster Analysis
0.01
1
0
600
6,000
molecules
60,000
600,000
www.fireflybio.com