BREAKINGSentieon ranks 1st in ALL THREE sub-challenges of the precisionFDA Brain Cancer Challenge DNAscope LongRead runs ~6x faster than DeepVariant on PacBio HiFi Cost per genome demonstrated below $1.50 on AMD EPYC DNAseq output matches GATK bit-for-bit Pangenotyper wins Bio-IT World 2024 Best-of-Show No GPU required - speed comes from algorithms, not hardware BREAKINGSentieon ranks 1st in ALL THREE sub-challenges of the precisionFDA Brain Cancer Challenge DNAscope LongRead runs ~6x faster than DeepVariant on PacBio HiFi Cost per genome demonstrated below $1.50 on AMD EPYC DNAseq output matches GATK bit-for-bit Pangenotyper wins Bio-IT World 2024 Best-of-Show No GPU required - speed comes from algorithms, not hardware
Sentieon logo
SENTIEON, INC. — MOUNTAIN VIEW / SAN JOSE, CA
Genomics · Precision Medicine · Est. 2014

Sentieon

The quiet Bay Area company whose variant-calling software wins the field's toughest accuracy contests - then runs on the CPUs you already own.

2014
FOUNDED
1st
precisionFDA × MANY
99.8%+
HiFi PRECISION & RECALL
<$1.50
COST PER GENOME

Somewhere right now, a sequencing machine is finishing a run. The DNA is done; the reading is done. What's left is the part nobody films for the brochure - the quiet, computational slog of turning billions of short text fragments into a list of the handful of places where this genome differs from every other. For years that step was the traffic jam of modern biology. The sequencer sprinted; the analysis crawled. Sentieon exists to erase that gap, and it has spent a decade doing it so effectively that most people never notice the software is there at all.

That is the strange thing about Sentieon. It is one of the most consequential names in genomics that the public has never heard of. Eleven-ish people, a modest office in the Bay Area, no splashy billboard campaign - and yet when the U.S. Food and Drug Administration runs its precisionFDA challenges to find who can call genetic variants most accurately, Sentieon keeps walking away with the top spots. It ranked first in all three sub-challenges of the Brain Cancer Predictive Modeling Challenge. It won most categories of the Truth Challenge V2 for the hardest-to-map regions of the genome. It led the ICGC-TCGA DREAM challenge for somatic mutations. The trophy shelf belongs to a company you could fit in a large elevator.

“Enable precision data for precision medicine.”

- SENTIEON'S MISSION, IN SIX WORDS

OriginEngineers who wandered in from a different field

Sentieon was founded in 2014 by Jun Ye, alongside co-founders Hanying Feng and Xiaofeng Liu. Their resumes are the first clue to why the company thinks differently. This was not a team of career geneticists. Their backgrounds run through image processing, telecommunications, computational lithography - the exacting math used to etch circuits onto silicon - and data mining. To that crowd, a genome is not sacred biology. It is a signal-processing problem: a very noisy channel, an enormous stream of data, and a demand for precision under real-world constraints.

That outsider's eye mattered. The reigning tool of the era was GATK, the Broad Institute's open-source workflow - the gold standard, and famously slow. Most companies would have tried to beat it by inventing a flashier method that produced different, arguably-better answers. Sentieon made a subtler bet. What if you could produce exactly the same answers - bit-for-bit identical to GATK - but many times faster? No retraining the field. No re-validating clinical pipelines. Just the same trusted result, delivered before the coffee got cold.

The ProductA toolbox that speaks every dialect of sequencing

What began as an ultrafast GATK alternative has grown into a full suite. The naming is engineer-plain, which is somehow reassuring: you always know what a Sentieon tool does.

DNAseq

Short-read germline calling for SNPs and indels that matches GATK output exactly - at a fraction of the runtime.

DNAscope

The precisionFDA award-winner. Uses platform-specific machine-learning models to push accuracy higher across SNPs, indels, SVs and CNVs.

DNAscope LongRead

Germline calling from PacBio HiFi and Nanopore. A 30x HiFi genome in under four hours - roughly 6x faster than DeepVariant.

DNAscope Hybrid

Fuses short and long reads from the same sample into one pipeline, so no signal is wasted.

TNseq / TNscope

Somatic tumor-normal calling with UMI consensus for cancer and liquid-biopsy work - SNVs, indels and structural variants.

DNAscope Pangenome

Aligns reads to graph reference genomes, catching variation that a single linear reference misses.

Above: the Sentieon stack, left to right - short reads, long reads, both at once, and the cancer toolkit. One company, every sequencing dialect.

The common thread is a refusal of the usual bargain. In genomics you were long told to pick two of three: fast, accurate, cheap. Sentieon's pitch is that clever algorithms dissolve the tradeoff. Everything runs on ordinary CPUs - the machines already sitting in the data center or the cloud account. No GPUs to buy, no FPGAs to program, no exotic accelerator to babysit. The speed is in the math, not the metal.

“Award-winning accuracy on any generic CPU-based computing system - no specialized hardware required.”

- THE SENTIEON DESIGN PRINCIPLE

Why It MattersWhat you can actually do with it

Strip away the acronyms and the value is human. A clinical lab running a cancer panel needs an answer today, not next week, because a patient is waiting. A pharmaceutical team screening thousands of samples needs the compute bill to stay sane. A researcher chasing a rare pediatric disease needs to trust that the variant on the screen is real and not an artifact of the software. Sentieon sits underneath all of them, doing the unglamorous middle step - secondary analysis - fast enough and accurately enough that it stops being the thing everyone worries about.

That is why its customers skew toward people who cannot afford to be wrong: pharma R&D, molecular diagnostics labs, sequencing-platform manufacturers who validate against it, and academic genomics centers. When the AMD benchmark showed DNAscope calling a full human genome for under $1.50 in compute, it wasn't a marketing stunt - it was a budget meeting getting shorter, a study getting bigger, a screen that used to be rationed becoming routine.

CultureReceipts, not adjectives

Sentieon markets the way engineers argue: with evidence. Where other firms reach for superlatives, Sentieon reaches for a benchmark, a peer-reviewed paper on bioRxiv, or a challenge result with an independent scoreboard. It is a company that would rather show you an F1 score than tell you it's the best. In a field crowded with claims, that restraint is itself a strategy - and it is a large part of why an eleven-person team gets taken seriously against organizations a hundred times its size.

The partnerships fit the pattern. A collaboration with Velsera produced the Pangenotyper graph-genome tool that won Bio-IT World's 2024 Best-of-Show. The pipelines are available across AWS, Azure and Google Cloud, and tuned with platform-specific models for PacBio, Nanopore, Illumina and Element. Sentieon doesn't try to own the whole stack. It tries to be the fast, trustworthy layer everyone else builds on.

The TimelineA decade of quiet wins

2014

Sentieon founded to rebuild genomics secondary analysis around speed and accuracy.

2015

Early seed funding; DNAseq establishes itself as the bit-for-bit-faster GATK alternative.

2021

Wins the precisionFDA Truth Challenge V2 and launches its PacBio long-read workflow.

2022

Publishes the DNAscope LongRead preprint: high-accuracy, fast germline calling from HiFi reads.

2024

Pangenotyper (with Velsera) wins Bio-IT World Best-of-Show; AMD benchmark shows sub-$1.50 genomes.

The Trophy Shelf

Marginalia

Five things that stick with you

Outsiders

The founders came from image processing, telecom and chip lithography - not biology. They treated DNA like a signal-processing problem.

Bit-for-bit

DNAseq matches GATK output exactly, so labs can switch pipelines without re-validating a single downstream result.

David vs. Goliath

An ~11-person team routinely out-ranks organizations a hundred times its size on public FDA leaderboards.

No accelerators

No GPUs, no FPGAs. The speed is entirely algorithmic - it runs on the hardware you already have.

CodaBack to that finishing sequencer

Return to the machine from the opening - the run just completed, the data waiting. In the old world, that was the moment the clock started ticking against you: queue the job, block the cluster, wait, hope the answer arrives before the meeting, before the deadline, before the patient's next appointment. Sentieon's whole reason for being is to make that moment boring. The reads go in; the trusted variants come out, fast enough that the analysis is no longer the thing you plan your week around.

That is the unshowy revolution. Not a cure, not a headline, but the removal of a bottleneck - repeated across thousands of labs, millions of samples, quietly compounding. The sequencer finishes. The answer is already almost there. Somewhere in the middle sits eleven people in the Bay Area who decided the waiting was optional, and then proved it.