Summarized four Grail's clinical trials project, all info collected from Grail publications and slides.

1. GRAIL

2. CCGA Is A Prospective Longitudinal Cohort Study
Blood samples
(from all participants)
Tissue samples
(cancer only)
15,000+ participants
70% with cancer
30% without
141 Active Sites
Targeted sequencingcfDNA, WBCs
Whole-genome sequencingcfDNA, WBCs
Targeted & whole-genome bisulfite sequencingcfDNA
Follow-up for 5 yrs
Vital status & cancer
status
Whole Genome Sequencingof tumor tissue
Participants with cancer:Data on treatment, recurrence,
mortality
Participants without cancer:Data on cancer diagnosis +
treatment, recurrence, mortality; or remain cancer-free
FPI: 05/2020; 15,254 enrolled;
Whole transcriptome sequencing cfRNA
cfDNA = Cell-Free Deoxyribonucleic Acid; WBC = White Blood Cell; cfRNA = Cell-Free RibonucleicAcid

3. Circulating Cell-free Genome Atlas Study - CCGA

4. Substudy 1: Comparison of assays to identify most
suitable assay for further development
• Discovery phase completed 20181
• Compared 3 sequencing approaches
• Targeted sequencing (SNVs)
• WGS (CNVs)
• WGBS (Methylation)
• Demonstrated feasibility of multi-cancer
detection with low false positive rate (~2%)
• Methylation-based assay selected for further
development
3 Substudies
1. Discovery
Training, n=1,785
Validation, n=1,015
2. Training/Validation
Training, n=3,133
Validation, n=1,354
3. Further Validation
~5,000 Participants
CCGA, Circulating Cell-free Genome Atlas study (NCT02889978); CNV, copy-number variation; SNV, single-nucleotide variant; WGBS, whole-genome bisulfite sequencing; WGS, whole-genome
sequencing.
1
Klein EA, et al. J Clin Oncol 2018;36(15_suppl):12021.

5. CCGA Sub study 1
2,800 Participants
Training Set (N=1,792) Test Set (N=1,008)
1,785 Clinically Locked
1,627 Clinically Evaluable
• 878 Cancer
• 580 Non-Cancer
• 169 Non-Cancer assay Controls
1,399 Analyzable with Assay Data
• 841 Cancer
• 437 with tumor tissue
• 558 Non-Cancer
• 52 (3%) excluded based on
eligibility criteria
• 106 (6%) excluded due to
missing stage
• 228 (13%) excluded due to
unevaluable assay data for
one or more assays
• 169 (10%) non-cancer assay
controls excluded 其他：肾癌，子宫癌，胰腺癌，食道癌，淋巴瘤，头颈
癌，卵巢癌，肝癌，黑色素瘤，宫颈癌，多发骨髓瘤，
白血病，甲状腺癌，膀胱癌，胃癌，肛直肠癌，其他原
发性癌症

6. CCGA Has Geographically Diverse Enrollment
Representative of United States Population
Active/Enrolling Training Test Training and Test
142 active sites
representing 24
states in the U.S.
and one site in
Canada

7. Prototype Sequencing Assays Generate Signals
Used by Classifiers for Cancer vs Non-Cancer
All Major Somatic and Epigenetic cfDNA Features Characterized

8. Classification Scores Across Assays Are Highly Correlated and
Proportional to Circulating Tumor DNA Fraction (ctDNA)
Stage
I/II
Stage
III/IV
Stage
I/II
Stage
III/IV
13
34
Cases
Colorectal Cancer Subset
Targeted WGS WGBS
Sensitivity at95% Specificity
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

9. High Biological Signal in Typically Unscreened Cancers
• Subset analysis of 196 CCGA cases with cancer diagnosis associatedwith 5-year
cancer-specific mortality of >50%1
○ Includes lung (118), ovarian (16), pancreatic (25), hepatobiliary (13), and esophageal
(24) cancers
Stage I/II/III
Stage IV
117
79
Cases
1
5-year cancer-specific mortality rates for persons aged 50-79 from SEER18, 2010-2014; https://seer.cancer.gov.
Targeted WGS WGBS
Sensitivity at 95% Specificity
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

10. Substudy 2: Training and Validation of a
Targeted Methylation Assay
• Training and initial validation completed
20201
• A targeted methylation assay was developed
developed targeting key informative
methylation regions
• Training and validation of a targeted
methylation classifier
3 Substudies
1. Discovery
Training, n=1,785
Validation, n=1,015
2. Training/Validation
Training, n=3,133
Validation, n=1,354
3. Further Validation
~5,000 Participants
CCGA, Circulating Cell-free Genome Atlas study (NCT02889978).
1
Liu MC, et al. Ann Oncol. 2020;31(6):745-759.

11. Target Selection Using a Machine Learning Algorithm
4,000 CCGA Cancer
cfDNA Methylation Sequence Data
(+1,000 Tissues)
50+ Cancer Types
Early and Late Stage
2,000 CCGA Non-Cancer
cfDNA Methylation Sequence Data
Matched on age distribution by
gender in CCGA Substudy 1 discovery
cohort
Non-Cancer Conditions: Metabolic,
Hematological, Inflammatory, Auto-
immune
Public Sequence Data
The Cancer Genome Atlas
Oncoviruses
Machine Learning Algorithm
Lung
Colon
Non-Cancer
Lung
Lung
Non-Cancer
Cancer
Tissue Of
Origin

13. Substudy 2: High Clinical Suspicion (HCS)1 of
Cancer Subgroup Analysis
Participants
• Confirmed pathologic diagnosis of cancer,
OR
• High suspicion of diagnosis of cancer
Objective
• Evaluate test performance in participants
with HCS of cancer but without a conformed
conformed pathologic diagnosis at the time
of enrollment
3 Substudies
1. Discovery
Training, n=1,785
Validation, n=1,015
2. Training/Validation
Training, n=3,133
Validation, n=1,354
3. Further Validation
~5,000 Participants
1
High suspicion for a cancer diagnosis by clinical and/or radiological assessment, with planned biopsy or surgical resection to establish a definitive diagnosis within 6 weeks (42 days) after study blood
draw.
HCS subgroup
Training, n=213
Validation, n=90

14. Substudy 2 performance testing – Participant
Disposition
CCGA Substudy 2
{N = 4841, 2836 cancer, 2005 non-cancer}
STRIVE
(N = 2202, all non cancer)
Analysis Population
Training, n = 3052 (1531 Cancer, 1521 Non-Cancer [892 non-cancer samples from STRIVE])
Validation, n = 1264 (654 Cancer, 610 Non-Cancer [337 non-cancer samples from STRIVE])
354 reserved for tissue
reference set
Training, n = 3,133 (1742 cancer, 1391 non-cancer)
Validation, n = 1354 (740 cancer, 614 non-cancer)
Clinically Locked and Evaluable
Training, n = 3032, (1654 Cancer, 1378 Non-Cancer)
Validation, n = 1316 (708 Cancer, 608 Non-Cancer)
Analyzable
Training, n = 3021, (1646 Cancer, 1375 Non-Cancer)
Validation, n = 1308 (703 Cancer, 605 Non-Cancer)
5 (< 1%) ineligible
5 (<1%) unlocked
13 (<1%) prior cancer dx/tx
78 (2%) unconfirmed
cancer/tx status
Excluded From Training
11 (<1%) non-evaluable
606 (20%) reserved for
future analysis
255 (8%) follow-up NA
Excluded From Validation Excluded From Training Excluded From Validation
Training, n = 1587
Validation, n = 615
Clinically Locked and Evalable
Training, n = 1460
Validation, n = 592
Analyzable
Training, n = 1460
Validation, n = 592
2 (<1%) unlocked
1 (<1%) prior cancer dx/tx
28 (2.1%) unconfirmed
cancer/tx status
8 (<1%) non-evaluable
606 (20%) reserved for
future analysis
255 (8%) follow-up NA
2 (<1%) ineligible
86 (5%) unlocked
38 (2%) presence or
suspicion of cancer
318 (22%) prior cancer
history confirmed or
unknown 250 (17%) F/U not
available
5 (<1%) ineligible/not
evaluable
9 (1.5%) clinically unlocked
9 (1.5%) presence or
suspicion of cancer
2 (<1%) non-evaluable
152 (26%) prior cancer
history confirmed or
unknown
101 (17%) F/U not available

15. Substudy 2: Comparable Demographics in
Cancer and Non-Cancer Cohorts
BMI, body-mass index; SD, standard deviation.
1:Cancer status was derived per statistical analysis plan (ie, was confirmed). Cancers by training/validation: anus (13/5), bladder (11/4), breast (247/104), cervix (11/7), colon/rectum (122/53), esophagus (50/21), gallbladder
(11/3), head and neck (62/25), kidney (56/25), liver/bile duct (29/11),
lung (260/111), lymphoid leukemia (38/19), lymphoma (109/50), melanoma (7/1), myeloid neoplasm (4/1), ovary (37/17), pancreas (84/39), plasma cell neoplasm (34/12), prostate (188/84), sarcoma (17/5), stomach (17/8),
thyroid (4/1), urothelial tract (5/0), uterus (84/36), and other (31/12);
“other”: in training included brain (3), Merkel cell carcinoma (1), mesothelioma (6), penis (1), pleura (1), small intestine (3), testis (1), thymus (2), unspecified (1), urethra (1), vagina (1), vulva (9); and in validation included
brain (1), orbit (1), penis (1), skin cancer (not basal cell carcinoma,
squamous cell carcinoma, or melanoma) (3), small intestine (1), testis seminoma (1), thymus (1), vagina (1), and vulva (2). 2Includes Asian, Native Hawaiian, or Pacific Islander; American Indian, or Alaska Native; Other;
Missing.

25. Substudy 3: Further Assay Refinement and
Clinical Validation
Participants
• Confirmed pathologic diagnosis of cancer,
OR
• High suspicion of diagnosis of cancer
Objective
• Evaluate test performance in participants
with HCS of cancer but without a conformed
conformed pathologic diagnosis at the time
of enrollment
3 Substudies
1. Discovery
Training, n=1,785
Validation, n=1,015
2. Training/Validation
Training, n=3,133
Validation, n=1,354
3. Further Validation
~5,000 Participants
1
High suspicion for a cancer diagnosis by clinical and/or radiological assessment, with planned biopsy or surgical resection to establish a definitive diagnosis within 6 weeks (42 days) after study blood
draw.
HCS subgroup
Training, n=213
Validation, n=90

26. Substudy 3: Demographics and Baseline
Characteristics

27. Cancer Signal Detection: Specificity, Overall
Sensitivity, and Signal Origin Prediction

28. Sensitivity of Cancer Signal Detection by
Clinical Stage

29. Sensitivity of Cancer Signal Detection in 12 Pre-Specified
Cancers Responsible for Two-Thirds of Cancer Deaths

30. Sensitivity of Cancer Signal Detection by
Cancer Class

31. Sensitivity of Cancer Signal Detection in
Cancers With and Without Common Screening

32. Specificity, Sensitivity, and Signal Origin
Prediction by Age Group

33. Introduction - MCED

34. Overview of MCED test workflow

35. Definitions of key metrics for measuring sample
quality and tumor cfDNA abundance

36. MCED test demonstrated high analytical
sensitivity and specificity

37. Accurate cancer signal detection and consistent cancer signal
origin predictions across allowable input range

38. MCED test outputs were concordant in
within-run and between-run sample pairs

39. Binary target coverage was consistent across runs
for both cancer and non-cancer samples

40. The MCED test workflow was not susceptible
to sample cross-contamination

41. Investigated interferents did not affect cfDNA yield,
binary target coverage, or test results

42. Conclusions