> For the complete documentation index, see [llms.txt](https://help.dragen.illumina.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.dragen.illumina.com/dragen-v4.5/product-guides/dragen-v4.5/germline/standard-sbs/small_variants_accuracy.md).
# WGS Small Variants
Generated on **2026-05-07**
## Small variants
The DRAGEN Germline Small Variant Caller takes mapped and aligned DNA reads as input and calls SNPs and indels through a combination of column-wise detection and local de novo assembly of haplotypes. For more information, refer to the [DRAGEN user guide](/dragen-v4.5/product-guides/dragen-v4.5/dragen-dna-pipeline/small-variant-calling.md).
**DRAGEN**: DRAGEN 4.4.6 vs 4.5.4 | **Truthset**: GIAB/NIST high-confidence, GIAB/NIST v4.2.1 | **Reference**: GRCh38
Results reported for HG001-HG007 sequenced with NSX 1.3 with 10B flow cell downsampled at 35x raw coverage.
SNPs table (click to expand)
| Sample | Version | Recall | Precision | F1-score | FN | FP |
| ------ | ------- | ------ | --------- | -------- | ---- | ---- |
| HG001 | 4.4.6 | 1.000 | 0.999 | 0.999 | 1336 | 2634 |
| HG001 | 4.5.4 | 1.000 | 1.000 | 1.000 | 721 | 1658 |
| HG002 | 4.4.6 | 0.999 | 1.000 | 0.999 | 2896 | 1749 |
| HG002 | 4.5.4 | 0.999 | 1.000 | 1.000 | 2198 | 1026 |
| HG003 | 4.4.6 | 0.999 | 0.999 | 0.999 | 3511 | 3005 |
| HG003 | 4.5.4 | 0.999 | 0.999 | 0.999 | 2936 | 2338 |
| HG004 | 4.4.6 | 0.999 | 0.999 | 0.999 | 3748 | 2584 |
| HG004 | 4.5.4 | 0.999 | 1.000 | 0.999 | 2963 | 1803 |
| HG005 | 4.4.6 | 0.999 | 0.999 | 0.999 | 3023 | 2047 |
| HG005 | 4.5.4 | 0.999 | 1.000 | 0.999 | 2413 | 1369 |
| HG006 | 4.4.6 | 0.999 | 0.999 | 0.999 | 3103 | 2408 |
| HG006 | 4.5.4 | 0.999 | 1.000 | 0.999 | 2484 | 1699 |
| HG007 | 4.4.6 | 0.999 | 0.999 | 0.999 | 3467 | 2327 |
| HG007 | 4.5.4 | 0.999 | 1.000 | 0.999 | 2955 | 1602 |
Indels table (click to expand)
| Sample | Version | Recall | Precision | F1-score | FN | FP |
| ------ | ------- | ------ | --------- | -------- | ---- | ---- |
| HG001 | 4.4.6 | 0.996 | 0.997 | 0.997 | 1867 | 1437 |
| HG001 | 4.5.4 | 0.996 | 0.997 | 0.997 | 1849 | 1231 |
| HG002 | 4.4.6 | 0.996 | 0.997 | 0.997 | 2101 | 1421 |
| HG002 | 4.5.4 | 0.996 | 0.998 | 0.997 | 1998 | 1229 |
| HG003 | 4.4.6 | 0.996 | 0.997 | 0.997 | 1852 | 1420 |
| HG003 | 4.5.4 | 0.997 | 0.998 | 0.997 | 1789 | 1164 |
| HG004 | 4.4.6 | 0.996 | 0.997 | 0.997 | 1898 | 1384 |
| HG004 | 4.5.4 | 0.996 | 0.998 | 0.997 | 1877 | 1226 |
| HG005 | 4.4.6 | 0.998 | 0.998 | 0.998 | 897 | 670 |
| HG005 | 4.5.4 | 0.998 | 0.999 | 0.998 | 810 | 554 |
| HG006 | 4.4.6 | 0.997 | 0.998 | 0.998 | 1147 | 838 |
| HG006 | 4.5.4 | 0.997 | 0.998 | 0.998 | 1116 | 741 |
| HG007 | 4.4.6 | 0.997 | 0.998 | 0.998 | 1246 | 868 |
| HG007 | 4.5.4 | 0.997 | 0.998 | 0.998 | 1161 | 762 |
**HG002 T2TQ100 performance**
**DRAGEN**: 4.4.6 vs 4.5.4 | **Truthset**: T2TQ100-v1.1-v0.019 | **Reference**: GRCh38
HG002 T2TQ100-v1.1-v0.019 small variants comparison (click to expand)
| Version | SNPs FN | SNPs FP | SNPs FP+FN | Indels FN | Indels FP | Indels FP+FN |
| ------- | ------- | ------- | ---------- | --------- | --------- | ------------ |
| 4.4.6 | 21571 | 8752 | 30323 | 40208 | 23681 | 63889 |
| 4.5.4 | 16490 | 5764 | 22254 | 36945 | 19661 | 56606 |
**A legacy of continuous innovation and improving accuracy in each version.**
Historical HG002 benchmarked releases using the HG002 small variants NIST v4.2.1 truthset show FP+FN dropping from 41,244 in DRAGEN 3.4.5 to 4,364 in DRAGEN 4.5.4.
Historical version trend table (click to expand)
| Version | FP+FN |
| ------- | ------ |
| 3.4.5 | 41,244 |
| 3.5.7 | 40,056 |
| 3.6.3 | 39,046 |
| 3.7.5 | 21,479 |
| 3.8.4 | 20,402 |
| 3.9.3 | 19,052 |
| 3.10.4 | 12,851 |
| 4.0.3 | 12,815 |
| 4.2.4 | 11,163 |
| 4.3.3 | 6,615 |
| 4.4.6 | 5,955 |
| 4.5.4 | 4,364 |
**Mosaic detection**
Non-cancer post-zygotic mosaic variants have typical allele fraction (AFs) lower than 50% and therefore more challenging to find with the default small variant caller that has been optimized to detect germline variants with typical AFs of 0%, 50% or 100%. The Mosaic variant caller has been designed to address this gap by training a dedicated machine learning model trained on lowAF variants. For more information, refer to the [DRAGEN user guide](/dragen-v4.5/product-guides/dragen-v4.5/dragen-dna-pipeline/small-variant-calling/mosaic-detection.md).
**DRAGEN**: DRAGEN 4.5.4 vs 4.4.6 | **Truthset**: HG002 NIST Mosaic v1.0 | **Reference**: GRCh38 Results reported for HG002 sample [fastqs](https://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/data/AshkenazimTrio/HG002_NA24385_son/NIST_HiSeq_HG002_Homogeneity-10953946/HG002_HiSeq300x_fastq/) sequenced with HiSeq2500 at 300x available from NIST.
HG002 mosaic small variants comparison (click to expand)
| Version | SNPs FN | SNPs FP | SNPs TP | SNPs FP+FN | Indels FN | Indels FP | Indels TP | Indels FP+FN |
| ------- | ------- | ------- | ------- | ---------- | --------- | --------- | --------- | ------------ |
| 4.4.6 | 0 | 318 | 85 | 318 | 0 | 196 | 0 | 196 |
| 4.5.4 | 0 | 311 | 85 | 311 | 0 | 185 | 0 | 185 |
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://help.dragen.illumina.com/dragen-v4.5/product-guides/dragen-v4.5/germline/standard-sbs/small_variants_accuracy.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.