TABLE 2.

Intended use performance requirements for spatial biology using mIF and MSI.

Translational attributes of mIF (PhenoImager assay and workflow) for biopharma discovery/exploratory trials, clinical research/trial inclusion and standard-of-care/diagnostics.

Exploratory		Trial inclusion	Diagnostic
Early Phase (I/II)	Example data use	Trial Inclusion (I/II/III)	Therapy Decisions
Hypothesis Generation, incl. MOA		Cohort Expansion	Transfer to IVD Partner
Profiling—Biomarker Prevalence		Regulatory Submission	Regulatory Submission
RUO-level validation (may be CAP/CLIA or GCLP1 levels)	Requirements	CAP-level validation	LDT: CAP and CLIA levels
		CLIA registration	IVD: FDA clearance (PMA, ssPMA)
		GCLP a , ISO 15189	GCLP a , GMP, ISO 13485
Antibody, Specificity/Accuracy b
Yes	Cell lines (endog, transfected) c	Yes
Optional	sh/siRNA or CRISPR c	Preferred	Yes
Optional	Peptide/protein competition c	Optional	Yes
Yes	RNA ISH	Yes
Optional	Orthogonal (e.g., WB, RT-PCR)	Preferred	Yes
Optional	Normal TMA d	Yes
Can proceed if “NO”	Antibody, freedom-to-operate	Can proceed if “NO”	Must be “YES”
mIF Analytic Development
Same	Mono-mIF concordance e	Qualitatively/quantitatively comparing staining of each individual target as a monoplex (DAB and IF) to each same target in mIF panel using serially adjacent tissue sections
Same	Drop Test and Interference f	mIF panel drop controls to determine antibody and/or fluorophore interference if markers are co-expressed using at least 2 different tissue samples and 8* serial sections of each (*for 6 marker panel)
Same	Signal Intensity g	Normalized counts of 10–30 for all Opal fluorophores, except Opal Polaris 780, where the recommended range is 1–10 counts
Same	Dynamic Range h	Signal-to-background (SNR) ratio of 10+ for adjacent channels
Same	Opal Signal Balance i	Ratios of signals between neighboring channels of 3:1 or less
Same	Opal Cross-Talk j	Residual cross-talk of ≤1% to ensure minimal interference with IA
Analytic Performance (robustness)
Yes (3–5 slides in 1 run)	Precision (intra-run) k	Yes (3–5 slides in 1 run), tissue samples preferably Neg L M H or −/+
Yes (3–5 slides in 2 additional runs)	Reproducibility (inter-run) k	Yes (3–5 slides in 2 additional runs), tissue samples preferably Neg L M H or −/+
10–20	Sensitivity/Prevalence (sample#)	40–50 m
Contextual (assay type), at least 2 operators and 1 instrument)	Inter-instrument/user	Contextual (assay type), at least 2 operators and 1 instrument)	Contextual (study type), at least 3 operators and 3 instruments)
Optional	Epitope stability	Preferred (from at least 3 timepoints, and up to 6 months	Yes (from at least 3 timepoints, and up to 12-18 months)
No	Reagent stability	Optional (Yes, if diluted RTU and stored)	Yes (real-time, accelerated and extended stability)
Same	Photostability l	Scanned repeatedly over the course of 6 months with <10% loss of signal
Preferred	Antibody Lot/Lot QC	Yes (ISO standard quality)	Yes (at least 3 cGMP Lots)
Can proceed if “NO”	Antibody, freedom-to-operate	Can proceed if “NO”	Must be “YES”
Algorithm Performance and Report
Fit-for-purpose algorithm (for use only with sample set/specific indication used during the development of algorithm)	Example data use	Fit-for-purpose algorithm (for use only with sample set/specific indication used during the development of algorithm)	General purpose algorithm (broad enough to use with an independent validation set) or fit-for-purpose algorithm for a specific indication
Preferred	Pathology Input	Yes. Manual annotations of tumor region, exclusions, etc
Initial training on 5–10 samples, 3–5 FOV per sample. Performance confirmed by senior scientist and/or pathologist. Algorithm re-trained when needed	Training n	Initial training on 5–10 samples, 3–5 FOV per sample. Performance confirmed by a pathologist. Statistically powered validation test set. Comparison with pathologist annotation, review, and sign-off. After successful validation, if issues arise requiring retraining, re-validation is required
Quality control by visual review of WSI (≥10% fields processed) or FOV (≤20 per slide) being processed, including run batch stain controls	Image QC Process o	Quality control by rigorous quantitative and pathologist visual review of a subset of areas from WSI (for ≤6 markers) or FOV (typically for ≥7 markers) being processed, including run batch stain controls
Optional	Locked Algorithm	Preferred	Yes
Same	Readout	Cell phenotype (frequency, density, proximity, compartment), upset plots, tissue maps
No	Report incl. Sign-Out	Yes. Pathologist reviews and approves/flags image analysis results

Assumptions and Notes

^a Notes: Good Clinical Laboratory Practice (GCLP) is a set of standards that provide guidance on implementing principles of Good Laboratory Practice (GLP) and directives within the ICH, Guideline for Good Clinical Practice (GCP) to the analysis of samples from a clinical trial. GLP is a quality system that covers the organizational process and the conditions under which studies are planned, performed, monitored, recorded, archived, and reported. GCP is an international ethical and scientific quality standard for designing, conducting, performing, monitoring, auditing, recording, analyzing, and reporting clinical trials that involve the participation of human subjects. By combining the GLP and GCP sets of guidelines, GCLP ensures the quality and reliability of the clinical trial data generated by laboratories.

^b Assumptions: Reviewer is a pathologist who is familiar with the target and can confirm the associated biology and staining pattern.

^c Assumptions: Cell lines can be identified and acquired showing endogenous expression or no expression of the target, or cell line can be transfected—preferably at different levels—with the target of interest, and/or with different but related members of the same protein family to further demonstrate specificity. Assumes also that target can be expressed as purified protein and sh/siRNA or CRISPR targeting is possible (cell line and/or gene).

^d Notes: Normal TMA, refers to a human tissue array used to test normal tissue specificity of antibodies that is designed in conformance with FDA guidelines and requirements for tissue cross-reactivity studies for IVD certification that contains 30+ types of normal organ, each single core case from at least three individuals.

^e Assumptions: Targets in the mIF, are paired with the appropriate Opal fluorophores e.g., the brightest Opal fluorophores with the weaker expressing proteins, and vice versa. Various correlation coefficients (e.g., Pearson, Kendall, Spearman, Lin, Interclass), ideally reporting above 0.9 or 90% positive concordance, are acceptable.

^f Notes: Each marker detection is not affected by the presence of other markers. Interference is established using a drop test to identify markers involved in an “umbrella effect,” a term commonly used to describe when a previously applied marker impedes the application of an additional marker that colocalizes with the first. This is commonly seen in mBF, assays if using DAB. With TSA, inhibition of antigen recognition can be of a mechanism other than steric hindrance (umbrella effect) and result from the depletion of sites for activated tyramide binding. Both are important to acknowledge in instances when using TSA, if ≥ 2 markers of interest are in the same cellular compartment.

^g Notes: These ranges support reliable and accurate data analysis. Reliable data can still be obtained when signals are as low as a few counts or as high as 50 or more counts, but risks are higher for cross talk issues.

^h Notes: SNR calculated by dividing the average signal intensity of the top brightest 20 cells by the average signal intensity of the weakest 10% of cells. Typical ratios are 100+ with high-performing antibodies; SNR, above three can still provide analytical value.

ⁱ Notes: Most relevant for using Opal fluorophores in the 520, 540, and 570 channels in the same panel (classic: Opals 520, 540, 570, 620, 650, and 690). With MOTiF, which replaces Opal 540 with Opal 480 and Opal 650 with Opal 780, the 6 fluorophores are more spectrally distinct.

^j Notes: There are two main sources of crosstalk; 1) instrumental crosstalk occurring when fluorescence signals leak from one channel to another due to imperfect filter optics or from inadequate crosstalk compensation algorithms; 2) staining crosstalk from actual fluorophore inaccurately labeling proteins on the sample resulting in residual fluorophores inadvertently binding to epitopes intended to be labeled by another fluorophore. It is very important to distinguish the two causes because resolving each is a different process.

^k Assumptions: Co-efficient of variation (CV) of 20% or less, preferably an upper %CV limit of 15%.

^l Assumptions: Slide stored at room temperature in dark. Camera exposure times should ideally be in the msec range for each fluorophore. This signal level translates to slide scan times of approximately 15 min for a 7-color assay at Magn. ×20 (∼0.5 µm × ∼0.5 µm pixel size) for a typical resection biopsy with area of 1.5 cm × 1.5 cm.

^m Notes: Capacity to detect cells with full range of marker expression. For predictive marker assays, labs should test a minimum of 20 positive cases and 20 negative cases. If the Lab Medical Director decides that a validation set of ≤40 cases is sufficient, they will need to document the rationale.

ⁿ Notes: Accuracy to count cells should be based on single markers. Cell classification accuracy using mIF, according to any marker co-expression, should be as provided by other single-cell platforms, while ensuring no signal contamination from neighboring cells. Cell segmentation can be challenging, complicated and may not always be robust across different cell types and/or due to staining variability or non-uniformity.

^o Note: Some WSI,analysis algorithms do not process whole slide images as a monolithic image, but rather in a field-based fashion with data from multiple FOV reduced to a single output afterward.