Tempus Ai Inc v. Guardant Health Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Tempus AI, Inc. (Delaware)
  • Defendant: Guardant Health, Inc. (Delaware)
  • Plaintiff’s Counsel: Quinn Emanuel Urquhart & Sullivan, LLP
• Case Identification: 3:25-cv-00622, S.D. Cal., 03/14/2025
• Venue Allegations: Plaintiff alleges venue is proper in the Southern District of California because Defendant Guardant Health maintains a regular and established place of business in the district, including a CAP-accredited and CLIA-certified laboratory in San Diego where it employs over 100 people.
• Core Dispute: Plaintiff alleges that Defendant’s oncology testing and data analytics platforms infringe four patents related to structuring and analyzing clinical/genomic data and using AI for histopathology image analysis.
• Technical Context: The technology at issue involves using large-scale data analytics and artificial intelligence to integrate disparate genomic, clinical, and imaging data to advance personalized cancer diagnosis and treatment.
• Key Procedural History: The complaint highlights prosecution history for several of the asserted patents, noting that the USPTO found the claims patent-eligible over abstract idea rejections and novel over the prior art. The complaint also references prior litigation involving Guardant, including a 2023 jury verdict finding willful infringement against Guardant in a case brought by Twinstrand Biosciences, to allege a pattern of disregarding intellectual property rights.

Case Timeline

Date	Event
2018-12-31	Priority Date for ’097 Patent
2019-09-19	Priority Date for ’859 Patent and '839 Patent
2020-02-28	Priority Date for ’041 Patent
2020-06-01	Alleged launch of GuardantINFORM platform
2020-08-01	Alleged integration of Guardant360 CDx with GuardantINFORM
2020-10-15	Notice of Allowance for ’041 Patent
2020-12-11	Notice of Allowance for ’097 Patent
2021-03-23	Issue Date for ’041 Patent
2021-04-27	Issue Date for ’097 Patent
2022-01-01	Alleged introduction of Guardant INFINITY
2023-01-06	Terminal disclaimer filed for '859 Patent application
2023-01-25	Notice of Allowance for ’839 Patent
2023-05-02	Issue Date for ’859 Patent
2023-01-01	Alleged introduction of Guardant Galaxy
2024-01-01	Alleged introduction of Guardant360 TissueNext
2024-10-08	Issue Date for ’839 Patent
2025-03-14	Complaint Filing Date

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 11,640,859 - Data Based Cancer Research and Treatment Systems and Methods

The Invention Explained

• Problem Addressed: The patent addresses the challenge of making sense of vast amounts of disconnected and unstructured cancer data from different sources (e.g., clinical records, genomic sequences) (Compl. ¶¶27-29). It notes that such data is often incomplete or inconsistent, making it difficult to use for deriving insights into treatment efficacy (Compl. ¶27; '859 Patent, col. 4:2-15).
• The Patented Solution: The invention describes a system that ingests raw clinical and genomic data into a "semi-structured" first database (a "data lake"), processes or "shapes" it into a "system structured data" format in a second database (a "data vault"), and then creates "application optimized" subsets of data in a third database (a "data mart") for use by specific research or clinical applications (Compl. ¶24; '859 Patent, abstract; '859 Patent, col. 10:11-11:13). This tiered data architecture is designed to make disparate data types usable for analysis (Compl. ¶31).
• Technical Importance: This structured approach to unifying multi-modal data was designed to enable large-scale, data-driven research and personalized medicine applications that were previously impractical due to data fragmentation (Compl. ¶29).

Key Claims at a Glance

• The complaint asserts at least Independent Claim 1 (Compl. ¶100).
• The essential elements of Claim 1 are:
  • Storing a set of user application programs.
  • For a plurality of cancer patients, performing the steps of:
    • (a) obtaining clinical records data in original forms.
    • (b) storing the clinical records data in a semi-structured first database.
    • (c) for each patient, using a next generation genomic sequencer to generate genomic sequencing data.
    • (d) storing the sequencing data in the first database.
    • (e) shaping at least a subset of the first database data to generate system structured data.
    • (f) storing the system structured data in a second database.
    • (g) for each user application program, selecting an application specific subset of data from the second database and storing it in a structure optimized for the application program in a third database.
• The complaint reserves the right to assert other claims (Compl. ¶97).

U.S. Patent No. 12,112,839 - Data Based Cancer Research and Treatment Systems and Methods

The Invention Explained

• Problem Addressed: As a continuation of the '859 patent's application, the '839 patent addresses the same core problem of fragmented and unstructured clinical and genomic data hampering cancer research and personalized treatment (Compl. ¶27; '839 Patent, col. 5:15-6:6).
• The Patented Solution: The method claimed in the '839 patent is similar to the '859 patent but adds steps for processing the output of a user application. After a user application selects data from a database, the system receives a "respective genomic variant characterization" from that application for each patient and stores this new characterization in a third database ('839 Patent, abstract; Compl. ¶25). This creates a feedback loop where analytical outputs are reintegrated into the structured data ecosystem.
• Technical Importance: This invention enables an iterative analytical process, where insights generated by one tool can be captured, stored, and used as inputs for subsequent analyses by the same or different tools (Compl. ¶33).

Key Claims at a Glance

• The complaint asserts at least Independent Claim 1 (Compl. ¶106).
• The essential elements of Claim 1 are:
  • Storing a set of user application programs.
  • For a plurality of subjects with cancer, performing the steps of:
    • (a) obtaining clinical records data in original forms.
    • (b) for each subject, using a next generation genomic sequencer to generate genomic sequencing data.
    • (c) shaping at least a subset of the genomic sequencing data to generate system structured data.
    • (d) storing the system structured data in a first database.
    • (e) for each user application program:
      • (i) selecting the application specific subset of data from the first database.
      • (ii) storing the application specific subset of data in a structure optimized for application program interfacing in a second database.
      • (iii) receiving the respective genomic variant characterization from the user application program for each subject.
      • (iv) storing the respective genomic variant characterization received from the user application program for each subject of the respective plurality of subjects in a third database.
• The complaint reserves the right to assert other claims (Compl. ¶97).

U.S. Patent No. 10,957,041 - Determining biomarkers from histopathology slide images

• Technology Synopsis: The patent addresses the inefficiencies of manual histopathology slide analysis by claiming a computer-implemented method for identifying biomarkers in a digital image of a tissue sample (Compl. ¶¶48; 50). The method involves separating the digital image into tiles and applying a "deep learning framework" comprising trained biomarker classification models to predict the presence of one or more biomarkers (Compl. ¶47).
• Asserted Claims: Independent Claim 1 (Compl. ¶112).
• Accused Features: The complaint alleges that the Guardant360 TissueNext PD-L1 test on the Guardant Galaxy platform infringes this patent (Compl. ¶95).

U.S. Patent No. 10,991,097 - Artificial intelligence segmentation of tissue images

• Technology Synopsis: The patent seeks to improve on conventional slide analysis by describing a method to create an "overlay map" on a digital image of a slide (Compl. ¶61). This involves receiving the digital image, separating it into tiles, identifying features of the tiles, and then identifying the "majority class of tissue visible" within each tile to determine a predicted class for the tile (Compl. ¶62).
• Asserted Claims: Independent Claim 1 (Compl. ¶118).
• Accused Features: The complaint alleges that the Guardant360 TissueNext PD-L1 test on the Guardant Galaxy platform infringes this patent (Compl. ¶96).

III. The Accused Instrumentality

Product Identification

• The accused products are Guardant’s platforms for oncology testing and data analysis, collectively identified as Guardant Galaxy, Guardant INFINITY, Guardant360 TissueNext, Guardant360 CDx, and GuardantINFORM (the "Accused Products") (Compl. ¶84).

Functionality and Market Context

• The complaint alleges these products form an integrated ecosystem that emulates Tempus's technology (Compl. ¶75).
• GuardantINFORM is described as a "text mining and natural language processing (NLP) precision oncology platform" that extracts and normalizes information from clinical documents and genomic data (Compl. ¶78). It allegedly integrates information from NGS tests like Guardant360 CDx and TissueNext (Compl. ¶87).
• Guardant360 TissueNext is an "analytically validated comprehensive next-generation sequencing panel" for solid tumors (Compl. ¶83). An AI-powered version of its PD-L1 test runs on the Guardant Galaxy platform (Compl. ¶82).
• Guardant Galaxy is described as a "suite of advanced analytical technologies" and an "AI-backed digital pathology platform" that enhances Guardant's cancer tests (Compl. ¶81; Compl. ¶82). A screenshot from a Guardant press release shows Guardant Galaxy will include "novel native AI analysis applications" (Compl. Ex. 6, p. 3).
• Guardant INFINITY is a "smart liquid biopsy platform" that powers Guardant's portfolio with "epigenomic insights" (Compl. ¶80).
• The complaint presents these products as Guardant's effort to compete with Tempus by moving into multi-modal data and AI-enabled platforms (Compl. ¶¶2; 75).

IV. Analysis of Infringement Allegations

The complaint includes preliminary claim charts in its exhibits, which are summarized below.

11,640,859 Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
storing a set of user application programs wherein each of the programs requires an application specific subset of data to perform application processes and generate user output;	Guardant’s Accused Products allegedly comprise a set of user application programs (e.g., TissueNext assay) that perform specific analyses (e.g., predict biomarkers) and generate user output.	¶102; Ex. 5, p. 6	col. 27:62-28:11
for each of a plurality of patients that have cancerous cells and that receive cancer treatment:	The Accused Products are used for patients with cancer (e.g., non-small-cell lung cancer) to select which treatment to receive.	¶102; Ex. 5, p. 19	col. 2:25-34
(a) obtaining clinical records data in original forms where the clinical records data includes cancer state information, treatment types and treatment efficacy information;	Guardant allegedly obtains clinical records through Test Requisition Forms and populates its LIMS system with patient-specific information. A screenshot of a Guardant LIMS report shows "patient-specific information" is included (Compl. Ex. 5, p. 22).	¶102; Ex. 5, p. 21	col. 22:3-15
(b) storing the clinical records data in a semi-structured first database;	The Accused Products allegedly store the obtained clinical records data in a "first database," with the GuardantINFORM platform integrating information from various sources.	¶102; Ex. 5, p. 24	col. 24:3-10
(c) for each patient, using a next generation genomic sequencer to generate genomic sequencing data for the patient’s cancerous cells and normal cells;	The Guardant360 TissueNext assay is alleged to use a next generation sequencer to generate genomic sequencing data from patient tissue biopsies (cancerous cells) and blood biopsies (normal cells).	¶102; Ex. 5, p. 27	col. 36:1-12
(d) storing the sequencing data in the first database;	The sequencing data generated is allegedly stored in a "first database," such as the GuardantINFORM data platform that "integrates information from NGS".	¶102; Ex. 5, p. 33	col. 32:20-25
(e) shaping at least a subset of the first database data to generate system structured data including clinical record data and sequencing data wherein the system structured data is optimized for searching;	Guardant's software (e.g., Guardant360 CDx Software) is alleged to use a "custom-developed analysis bioinformatics pipeline" to partition and process raw sequencing data into structured data for analysis.	¶102; Ex. 5, p. 37	col. 24:11-25
(f) storing the system structured data in a second database;	Guardant's information technology systems are alleged to "store a wide variety of information critical to our business," which includes a "bioinformatics pipeline" that includes a database for storing system structured data.	¶102; Ex. 5, p. 43	col. 24:26-30
(g) for each user application program: (i) selecting the application specific subset of data from the second database; and (ii) storing the application specific subset of data in a structure optimized for application program interfacing in a third database.	Guardant's platforms (e.g., Guardant Galaxy) are alleged to select application-specific subsets of data (e.g., pathology data) from a second database and store it in a structure optimized for interfacing in a third database.	¶102; Ex. 5, p. 46; Ex. 5, p. 51	col. 24:31-41

12,112,839 Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
storing a set of user application programs wherein each of the programs requires an application specific subset of data to perform application processes and generates a respective genomic variant characterization;	The Guardant360 TissueNext assay is alleged to be a user application program that requires a subset of data to perform analysis (e.g., biomarker prediction) and generates a "genomic variant characterization."	¶108; Ex. 6, p. 6	col. 60:16-25
for each of a plurality of subjects that have cancerous cells and that receive cancer treatment:	The Accused Products are used for a plurality of subjects with cancer to inform treatment decisions. A Guardant leaflet screenshot states its test is for "patients with advanced solid tumors" (Compl. Ex. 6, p. 17).	¶108; Ex. 6, p. 16	col. 4:16-24
(a) obtaining clinical records data in original forms where the clinical records data includes cancer state information, treatment types and treatment efficacy information;	Guardant allegedly obtains clinical records via a Test Requisition Form which includes fields for cancer stage and therapy, and its systems generate "results reports" populated with patient-specific information.	¶108; Ex. 6, p. 19	col. 117:1-12
(b) for each subject, using a next generation genomic sequencer to generate genomic sequencing data for the subject’s cancerous cells and normal cells;	The Guardant360 TissueNext assay allegedly uses a next-generation sequencer on tissue biopsies (cancerous cells) and may be ordered with a liquid biopsy for non-cancerous cell DNA.	¶108; Ex. 6, p. 22	col. 118:13-20
(c) shaping at least a subset of the genomic sequencing data to generate system structured data;	Guardant's software allegedly uses a bioinformatics pipeline that partitions raw data, executes "proprietary algorithmic reconstruction," and performs an alignment process to generate structured data.	¶108; Ex. 6, p. 28	col. 114:1-14
(d) storing the system structured data in a first database;	Guardant's bioinformatics pipeline is alleged to include a database for storing structured sequencing data, and its annual report states it manages data on-site and in cloud-based data centers.	¶108; Ex. 6, p. 31	col. 114:15-18
(e) for each user application program: (i) selecting the application specific subset of data from the first database; ... (iv) storing the respective genomic variant characterization ... in a third database.	The Accused Products are alleged to select data from a database for use in applications like the TissueNext assay. The resulting "genomic variant characterization" is then allegedly received and stored in a third database, such as the GuardantINFORM platform.	¶108; Ex. 6, p. 34; Ex. 6, p. 38; Ex. 6, p. 43; Ex. 6, p. 46	col. 582:1-20

Identified Points of Contention

• Scope Questions: A central question for the '859 and '839 patents will be whether Guardant's data architecture maps onto the specific "first," "second," and "third" database structure recited in the claims. The defense may argue its system, which integrates multiple services and data sources, does not have the discrete, sequentially optimized databases as claimed.
• Technical Questions: The infringement allegations rely on interpreting Guardant's marketing materials and high-level technical descriptions. A key question will be whether the underlying technical operation of Guardant's NLP, bioinformatics pipelines, and AI tools actually performs the specific steps of "shaping," "selecting," and "storing" as defined by the patents' specifications.

V. Key Claim Terms for Construction

• The Term: "semi-structured first database" (from '859 Patent, Claim 1)

• Context and Importance: This term defines the initial repository for disparate data types. Its construction is critical because it sets the baseline for the entire claimed data structuring process. Whether Guardant's initial data aggregation platform qualifies as "semi-structured" will be a key point of dispute.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The specification suggests this can be a general "data lake" for storing "un-structured, semi-structured, and/or structured data" from various sources, which may support a broad definition covering any large, heterogeneous data store ('859 Patent, col. 23:2-5; '859 Patent, col. 24:3-10).
  • Evidence for a Narrower Interpretation: The description of shaping data "from" the first database to create structured data "in" a second database could imply the first database is primarily for raw or less-structured data, potentially narrowing the scope to exclude systems that perform significant structuring upon ingestion ('859 Patent, col. 24:11-25).

• The Term: "genomic variant characterization" (from '839 Patent, Claim 1)

• Context and Importance: This term defines the analytical output that is fed back into the system's database structure. The scope of this term will determine what kind of analytical result from a "user application program" satisfies this limitation.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The patent provides a broad list of what this can include, such as "single nucleotide variants (SNVs), indels, CNV, etc.," as well as "variants of unknown significance" and "therapy match results," suggesting any output that characterizes a genomic feature may suffice ('839 Patent, col. 108:18-31).
  • Evidence for a Narrower Interpretation: Embodiments describe this characterization in the context of specific micro-services performing defined tasks like "variants to therapy relevance" or "TMB calculation" ('839 Patent, col. 26:19-29). A defendant might argue the term requires a specific type of classification output from a discrete application, not just any processed genomic data.

VI. Other Allegations

• Indirect Infringement: The complaint does not contain separate counts for indirect infringement under 35 U.S.C. § 271(b) or (c). While it alleges Guardant's products are used by third-party clinicians (Compl. ¶86), it does not plead specific facts to support the knowledge and intent elements required for inducement or contributory infringement, such as alleging Guardant instructs users to perform the claimed methods.
• Willful Infringement: The complaint alleges that Guardant had knowledge of the patents-in-suit at least as of the filing of the complaint (Compl. ¶¶101; 107; 113; 119). It further alleges a history of "disrespecting intellectual property rights" by pointing to a prior willful infringement verdict against Guardant and alleges Guardant intentionally copied Tempus's technology, which may support a claim for willful infringement (Compl. ¶¶75-76).

VII. Analyst’s Conclusion: Key Questions for the Case

• A central issue will be one of architectural equivalence: does Guardant's integrated suite of data platforms and services implement the specific multi-stage database architecture (e.g., semi-structured lake -> structured vault -> application-specific mart) recited in the '859 and '839 patents, or is this a superficial mapping of concepts onto a fundamentally different system?
• A second key question will be one of functional specificity for the AI patents ('041 and '097): does Guardant's use of a third-party AI module for its TissueNext test perform the precise, multi-step methods of tile-based analysis, classification with a "multiscale deep learning framework," and tissue "segmentation" as claimed, or is there a material difference in the computational techniques employed?
• A foundational legal challenge will likely be one of patent eligibility under 35 U.S.C. § 101: can Tempus successfully argue that its claims—which involve collecting data, structuring it in databases, and applying computational models—are directed to a concrete, practical application and an improvement in computer functionality, rather than an ineligible abstract idea of data analysis? The complaint's emphasis on the examiner's allowance over § 101 rejections suggests Tempus anticipates this fight.