Search Results

The lon PGM™ Dx Instrument System is intended for targeted sequencing of human genomic DNA (gDNA) from peripheral whole-blood samples and DNA and RNA extracted from formalin-fixed, paraffin-embedded (FFPE) samples. The lon PGM™ Dx Instrument System is not intended for whole genome or de novo sequencing.

Device Description

The Ion PGM™ Dx System is used for detection of human variant sequences from DNA from whole blood samples or RNA and DNA from FFPE tissue samples. Detectable variants include substitutions, insertions, and deletions.

The Ion PGMTM Dx System consists of the following:

Ion OneTouch™ Dx Instrument
Ion OneTouch™ ES Dx Instrument
Ion OneTouch™ Rack Kit
Ion PGM™ Dx Chip Minifuge
Ion PGM™ Dx Sequencer
Ion PGMTM Wireless Scanner
DynaMag™ Dx Kit—Tube & Plate
Ion Torrent™ Server
Torrent Suite™ Dx Software

The Ion PGM™ Dx System is used in conjunction with the following kits:

Ion PGM™ Dx Library Kit
Ion OneTouch™ Dx Template Kit
Ion PGM™ Dx Sequencing Kit
Ion 318™ Dx Chip Kit

The system should be used only by professionals trained in laboratory techniques and procedures and in the use of the system.

AI/ML Overview

The provided text describes the acceptance criteria and the studies performed for the Ion PGM™ Dx System. Here's a structured breakdown of the requested information:

1. Table of Acceptance Criteria and Reported Device Performance

The acceptance criteria are not explicitly stated in a single section as "acceptance criteria," but rather derived from the "Special conditions statement for performance" for both gDNA from whole blood and DNA/RNA from FFPE samples. The "Reported Device Performance" is taken from the "Non-Clinical Performance Data" section.

Feature / Metric	Acceptance Criteria (from "Special Conditions")	Reported Device Performance (from "Non-Clinical Performance Data")
gDNA from Whole Blood (SVA Panel)
Sequencing output	> 0.7 gigabases	Not explicitly reported for gDNA from whole blood in the performance data, but indicated as "validated to deliver" in the special conditions.
Reads	> 4 million	Not explicitly reported for gDNA from whole blood in the performance data, but indicated as "validated to deliver" in the special conditions.
Read length	up to 200 base pairs	Not explicitly reported for gDNA from whole blood in the performance data, but indicated as "validated to deliver" in the special conditions.
Mean Raw Read Accuracy	99.0% when compared to hg19	Not explicitly reported for gDNA from whole blood in the performance data, but indicated as "validated to deliver" in the special conditions.
SNV Detection Reproducibility	100% reproducibility for 440 unique SNV positions	Not explicitly reported in the Non-Clinical Performance Data for gDNA from whole blood. This is stated as a system capability in the "Special conditions statement for performance derived from gDNA from whole blood."
Indel Detection Reproducibility	100% reproducibility for various insertion/deletion lengths (1-4 bp insertions, 1-14 bp deletions)	Not explicitly reported in the Non-Clinical Performance Data for gDNA from whole blood. This is stated as a system capability in the "Special conditions statement for performance derived from gDNA from whole blood."
HPT limitation	Variants in homoploymer tracts exceeding 8 bases called as no calls	Not applicable; this is a known limitation, not a performance metric to be achieved.
Min coverage for germline DNA	>30X	Not explicitly reported in performance data. (This is a recommended use parameter)
FFPE Samples (Representative Assay)
Sequencing output	> 0.7 gigabases	Not explicitly reported for a representative assay in the performance data, but indicated as "validated to deliver" in the special conditions.
Reads	> 3 million	Not explicitly reported for a representative assay in the performance data, but indicated as "validated to deliver" in the special conditions.
Read length	up to 141 base pairs	Not explicitly reported for a representative assay in the performance data, but indicated as "validated to deliver" in the special conditions.
PPA (excluding no calls)	Not explicitly stated as a minimum.	Variant: 98.5% (195/198)Bin: 97.2% (176/181)Sample: 96.9% (158/163)
NPA (excluding no calls)	Not explicitly stated as a minimum.	Variant: 100.0% (118,155/118,159)Bin: 99.8% (942/944)Sample: 98.4% (124/126)
OPA (excluding no calls)	Not explicitly stated as a minimum.	Variant: 100.0% (118,350/118,357)Bin: 99.4% (1,118/1,125)Sample: 97.6% (282/289)
Repeatability (DNA variants, excl. no calls)	≥97.5% (95% CI lower limit)	≥98.8% (95% CI lower limit of ≥97.5%)
Repeatability (RNA variants, excl. no calls)	Not explicitly stated as a minimum for individual RNA variant locations, but overall ≥87.5% for positive variant location.	94.4% for each RNA clinical variant location. (For a specific ROS1 RNA variant for Sample C, it was 87.5% with 95% CI lower bound of 61.7%).
Call Rate (DNA pos. variants, excl. no calls)	Not explicitly stated as a minimum.	Mean: 96.60%, Median: 97.10%
Call Rate (RNA pos. variants, excl. no calls)	Not explicitly stated as a minimum.	Mean: 94.80%, Median: 95.50%
Call Rate (WT DNA, neg. calls, excl. no calls)	Not explicitly stated as a minimum.	Mean: 96.10%, Median: 95.00%
Call Rate (WT RNA, neg. calls, excl. no calls)	Not explicitly stated as a minimum.	Mean: 99.30%, Median: 99.30%
Tissue Input (% meeting conc.)	98.3% (59/60) had DNA ≥0.83 ng/uL and RNA >1.43 ng/uL. (This is a study finding, implicitly demonstrating the ability to meet the given concentration requirements for the assay under specific tissue input conditions.)	98.3% (59/60) met the concentration requirements (DNA ≥0.83 ng/uL, RNA >1.43 ng/uL).
DNA/RNA Input (Positive Call Rate)	100% positive variant call rate within 5-15 ng input range for a representative assay.	100% positive variant call rate within the input range tested (5-15 ng), supporting the 10 ng specified input. For clinical samples, one CD74-ROS1 fusion variant showed 100% positive calls at all input combinations, while the other showed rates as low as 50% at specific high input combinations (attributed to likely operator error).
DNA/RNA Input (Negative Call Rate)	Not explicitly stated as a minimum.	>95% for all except 4 sample/input combinations; cases with <95% were due to no calls. For clinical samples, DNA variants showed >95% negative call rates. The second CD74-ROS1 clinical sample showed 100% negative call rates for all test conditions where it was expected to be wild type.
Interfering Substances	Positive/Negative/Overall concordance with control was 100% (excluding no calls) for most interferents. For hemoglobin, positive concordance was 100%, negative 99.99%, overall 99.99%. (These are the observed results which met the study's goal of demonstrating non-interference).	For most interferents (Paraffin, Xylene, Ethanol, Protease, Wash buffer): 100% positive, negative, and overall concordance with control (excluding no calls). For Hemoglobin: 100% positive concordance, 99.99% negative concordance, 99.99% overall concordance (excluding no calls).
Cross Contamination Rate	Not explicitly stated as a specific rate, but the study aims to evaluate cross-contamination.	False-positive rate at DNA variant locations: 0% (0/100). False-positive rate at RNA variant locations: 1.25% (1/80), attributed to likely sample cross-contamination.
Minimal coverage needed for FFPE calling	≥347X for SNV, MNV, deletion; ≥41X for fusion.	Not explicitly reported in performance data. (This is a recommended use parameter)

2. Sample size used for the test set and the data provenance

Accuracy Study (FFPE):
- Sample Size: 290 FFPE tumor samples.
- Data Provenance: The document implies these are clinical samples ("human specimens," "FFPE tumor samples") but does not specify country of origin. It is a retrospective study since variants were compared against "validated reference detection methods."
Sample Reproducibility Study (FFPE):
- Sample Size: 2 WT (Wild Type) samples and 10 variant-positive samples. Each sample tested 8 times at each of 4 sites, for a total of 32 replicates per sample.
- Data Provenance: Not explicitly stated (e.g., country of origin, retrospective/prospective). These appear to be characterized samples used in a prospective, controlled study.
Assay Reproducibility Study (FFPE):
- Sample Size: 18 DNA samples (6 plasmid/clinical DNA blends, 12 clinical DNA samples) and 9 RNA samples (1 WT, 8 with RNA variants). Each pre-extracted sample run in duplicate with 2 different reagent lots (3 sites) or 3 reagent lots (1 site), resulting in 72 test determinations per DNA sample and 144 per RNA sample, total of at least 1,296 sequencing reactions.
- Data Provenance: Mixtures of plasmid and clinical DNA/RNA. Not explicitly stated (e.g., country of origin, retrospective/prospective). This is a prospective, controlled study.
Tissue Input Study (FFPE):
- Sample Size: 60 slide-mounted FFPE samples (30 resection with >20% tumor, 15 resection with <20% to ≥10% tumor (macrodissected), 15 CNB samples).
- Data Provenance: Clinical FFPE samples. Not explicitly stated (e.g., country of origin, retrospective/prospective). Presumed retrospective collection for the purpose of testing input requirements.
DNA and RNA Input Study (FFPE):
- Sample Size: 8 cell-line samples (as FFPE sections) for blending. 540 individual DNA and RNA libraries tested. Additionally, 4 clinical samples (2 DNA variants, 2 CD74-ROS1 fusion variants).
- Data Provenance: Cell-line samples prepared as FFPE sections, and clinical FFPE samples. Not explicitly stated (e.g., country of origin, retrospective/prospective). This is a prospective, controlled study.
Interfering Substances Study (FFPE):
- Sample Size: 8 FFPE samples (1 WT, 7 mutants) with 6 replicates each.
- Data Provenance: Not explicitly stated (e.g., country of origin, retrospective/prospective). This is a prospective, controlled study using selected samples.
Cross Contamination Study (FFPE):
- Sample Size: 8 FFPE cell line samples. 100 DNA and 80 RNA data points analyzed.
- Data Provenance: FFPE cell line samples. Not explicitly stated (e.g., country of origin, retrospective/prospective). This is a prospective, controlled study.

3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts

The document does not specify the number or qualifications of experts used to establish ground truth for any of the studies.

4. Adjudication method for the test set

The document does not describe an adjudication method for the test set. For the accuracy study, variants were compared against "validated reference detection methods" (NGS assay, ROS1 FISH test), which implies pre-established ground truth, not an adjudication process during the study.

5. If a multi-reader multi-case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

There is no indication that a multi-reader multi-case (MRMC) comparative effectiveness study was done, nor is there any mention of human readers or AI assistance. This device is a high-throughput genomic sequencer, not an imaging device typically involving human readers for interpretation in the way an MRMC study would apply.

6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done

The performance studies described (Accuracy, Reproducibility, etc.) are evaluating the Ion PGM™ Dx System as a whole, which includes the instrument and its associated software (Torrent Suite™ Dx Software). The "Performance Data" sections appear to describe the standalone performance of the entire system (instrument + software) without human intervention in the interpretation of the calls as the primary performance metric. The study results (PPA, NPA, call rates) reflect the algorithm's performance in detecting variants.

7. The type of ground truth used (expert consensus, pathology, outcomes data, etc)

Accuracy Study (FFPE): Ground truth was established using "validated reference detection methods," specifically:
- A validated NGS assay for SNV and deletion hotspot variants.
- A ROS1 FISH reference test for ROS1 fusions.
Reproducibility Studies (Sample, Assay, DNA/RNA Input, Interfering Substances, Cross Contamination): Ground truth was based on:
- Characterized samples (WT samples, variant-positive samples, plasmid/clinical DNA blends, FFPE cell line samples). This implies a pre-existing understanding of the variants present in these samples, likely established through prior sequencing or genetic analysis, serving as the reference standard.

8. The sample size for the training set

The document does not provide a specific sample size for a "training set." The studies described are performance validation studies, not studies related to the development or training of an AI/ML algorithm. The Ion PGM™ Dx System measures hydrogen ions during sequencing; its algorithms for base calling and variant detection are typically deterministic or rule-based, or if they involve machine learning, the training details are not disclosed in this document.

9. How the ground truth for the training set was established

As no "training set" is described for an AI/ML context, this question is not directly applicable. If the underlying algorithms for variant calling involved machine learning, the method for establishing ground truth for their training is not provided in this document. The document describes validation against characterized samples and reference methods.

Ask a Question

Ask a specific question about this device

Page 1 of 1