CogniMunePhenArI
Phase I · Specific Aims

Three aims, one turnkey protocol for immune monitoring

To prove the feasibility of CogniMune, this new program sets out three aims to establish a comprehensive, turnkey protocol for immune monitoring in patients undergoing disease-modifying therapies for AD — with clearly defined milestones and Go/No-Go decision gates.

Timeline at a glance

1
Aim 1 · Year 1 · Q1–Q2
Design & validate the 40-color panel
2
Aim 2 · Year 1 · Q2–Q3
Find the markers that track ARIA risk
3
Aim 3 · Year 1 Q3 – Year 2 Q4
Automate analysis with OT fingerprints
Interactive timeline

Phase I schedule and decision gates

Select an aim or a gate for details. Aims overlap so that each starts as soon as its inputs are ready.

Aim 1Aim 2Aim 3 Go/No-Go gate Use Tab and the arrow keys to move between items.
Aim 1 · Year 1 · Q1–Q2

Panel design, optimization & CLSI H62 validation

Design, test and validate the 40-color panel on healthy-donor blood (Figure 3, Figure 4).

Table view
ItemPeriodDescription
Aim 1Year 1 · Q1–Q2Panel design, optimization & CLSI H62 validation
Aim 2Year 1 · Q2–Q3Longitudinal study & ARIA-relevant markers
Aim 3Year 1 Q3 – Year 2 Q4Automated pipelines & visualization tools
Gate 1End of Year 1 · Q2Panel validated per CLSI H62
Gate 2End of Year 1 · Q3Immune profile versus treatment assessed
Gate 3End of Year 2 · Q4Automated fingerprint characterization → Phase II
The Specific Aims

What each aim delivers

Aim titles and success criteria are quoted from the proposal.

1 Aim 1 Year 1 · Q1–Q2

Design and validate robust FC panels to characterize peripheral immune cell populations associated with AD progression and treatment response

Research strategy: Design, optimize, and validate robust FC panels to characterize peripheral immune cell populations in patients with AD.

  • Design, test and validate a 40-color immunophenotyping panel for a spectral flow cytometer, based on OMIP-109 with minor updates.
  • Viability dye to discriminate live and dead cells; tetramer staining to track antigen-specific T cells.
  • Optimize reagent concentrations on whole blood and PBMCs from healthy donors of a local blood bank.
  • Validate the optimized panel according to CLSI H62 (Figure 4), under SOPs for daily instrument monitoring and new reagent lots.
Success criteria

Success criteria will be based on well published outcomes for panel design and optimization and for validation of this panel based on CLSI H62 guidelines.

2 Aim 2 Year 1 · Q2–Q3

Determine the most relevant markers to monitor ARIA risk in AD patients receiving disease-modifying therapies (n=30) through longitudinal study as compared to non-ARIA control groups

Research strategy: Assess the relevance of the designed staining panel in distinguishing non-AD patients (n=30) from AD patients undergoing therapy (n=30) via a longitudinal study.

  • Screening at the IU School of Medicine Brain Health Program: amyloid positivity (PET or CSF) and APOE ε4 genotyping before the start of therapy (Figure 2).
  • Blood collected before initiation (BL, week 0), at treatment visits through week 16 and at week 52 (T26), until ARIA occurs or treatment is discontinued (Figure 6).
  • Controls (Figure 5) and untreated AD patients (n=30) to compare immune signatures.
  • Compositional frequency data analyzed on the log-ratio scale: the proposal's temporal centered log-ratio follows each phenotype relative to the patient's own typical level.
Success criteria

The success criteria will be to determine the correlation of the immune profile with the minimum dose required and how this impacts the relevant population for future studies, specifically focusing on the immune cell populations affected by treatment.

Go vs. No-Go: assess the correlation and differences between the immune profile and the treatment, and evaluate how this relationship evolves throughout the course of the treatment.

3 Aim 3 Year 1 Q3 – Year 2 Q4

Develop automated data processing pipelines and visualization tools to enable standardized analysis of immunophenotyping data across healthcare settings

Automated pipelines that turn cytometry readouts into graph-based immune fingerprints using Sinkhorn-regularized optimal transport, compared over time with graph edit and spectral distances.

  • Sub-Aim 3-1 · Data fingerprinting: low-resolution graph representations of each patient's FC readout, built with the Sinkhorn distance.
  • Graphs compared with graph edit distance (GED) and spectral distance — on labeled populations or on raw statistical clusters.
  • Sub-Aim 3-2 · Data management and health informatics: Amazon Web Services with HIPAA compliance through a Business Associate Agreement (BAA) for future commercial processing.
  • Graph-based visualization with layouts derived from phenotype similarity, for clinicians and cytometry practitioners.
Success criteria

Success criteria will be to establish automated functional connectivity fingerprint characterization of the immediate response of the AD patients.

Go/No-Go

Predefined decision gates

Go/No-Go criteria support structured decision-making with quantitative performance thresholds, risk assessment and mitigation strategies. A clinically viable diagnostic should rest on substantial effect sizes — methods that detect such large, consistent signals do not require large cohorts.

Gate 1 · End of Year 1 · Q2

Panel validated per CLSI H62

Go Proposal
The Figure 4 acceptance criteria are met — intra- and inter-assay precision (%CV ≤ 25/35), inter-operator variability (% difference ≤ 20) and sample stability (% change from baseline ≤ 20) — and biological variability is reported as found. The panel is then used to immunophenotype patients with AD, alongside relevant controls.
No-Go Portal default
Criteria not met: the panel is re-optimized (titration, fluorochrome assignment, quality control) and re-validated before patient samples are analyzed.

Portal default Gate 1 counts as met only when every evaluated population passes in all four precision and stability studies and biological variability has been reported.

Gate 2 · End of Year 1 · Q3

Immune profile versus treatment

Go Proposal
Assess the correlation and differences between the immune profile and the treatment, and how this relationship evolves throughout the course of treatment. Success: the correlation of the immune profile with the minimum dose required, focusing on the immune cell populations affected by treatment.

Proposal A clinically viable diagnostic should rest on substantial effect sizes; in the screening context this is often framed by the Z′-factor, where a required Z′ = 0.5 implies ≈ 12σ. Planned sample size: 30 treated patients and 30 matched controls.

No-Go Proposal
“If a strong signal is not observable with n=30 per group, that would raise serious doubts about clinical utility.” Here n=30 per group means 30 treated patients and 30 matched controls.

Portal default The research portal's Gate 2 metric is the best Z′ between ARIA and non-ARIA patients in an ARIA vs non-ARIA marker-ranking run (the pinned run if there is one, otherwise the latest saved run), with Z′ ≥ 0.5 as an editable default threshold. The ARIA and non-ARIA groups together make up the 30 treated patients, so each group is smaller than 30.

Gate 3 · End of Year 2 · Q4

Automated fingerprint characterization

Go Proposal
Automated functional connectivity fingerprint characterization of the immediate response of the AD patients is established — the groundwork for the Phase II cloud-based system.
No-Go Portal default
The team refines and, when needed, redesigns algorithms and interfaces before Phase II.

Portal default The research portal also checks that pipeline results reproduce on re-run (reproducibility hash).

“Proposal” marks criteria quoted or paraphrased from the SBIR Phase I proposal; “Portal default” marks No-Go actions and quantitative thresholds used by the CogniMune research portal — editable defaults, not criteria stated in the proposal. Gate timing is placed at the end of each aim's period in the proposal.

What does Z′ = 0.5 mean?

Illustrative
Non-ARIA patients ARIA patients Separation band Overlap of ±3σ bands
Z′=1−3(σ1+σ2)|μ1−μ2|
0.50

Z′ = 0.5 — the proposal's screening-context benchmark: with equal spreads the means are 12σ apart.

Per the proposal, in the screening context the Z′-factor frames effect size: a required Z′ = 0.5 implies ≈ 12·σ. It is an illustration of effect size, not a Go threshold defined in the proposal; the research portal uses Z′ ≥ 0.5 between ARIA and non-ARIA patients as an editable default. Distributions here are schematic, not study data.

Study design

Who takes part

AD patients starting anti-amyloid therapy are screened at the Indiana University School of Medicine Brain Health Program; control samples come from Our Blood Institute and aging participants.

Screening workflow at IU School of Medicine · Figure 2

Cognitive screening

Patients screened based on cognitive test result

Amyloid positivity

PET scan or CSF

APOE ε4 genotyping

Carrier status determined

APOE ε4 carrier→ start of therapy
APOE ε4 non-carrier→ start of therapy

Cohorts · Figure 5 controls plus treated and untreated AD

170 participants planned across 5 cohorts
30nTreated · longitudinal

AD patients on anti-amyloid therapy

Screened at the IU School of Medicine Brain Health Program and starting lecanemab or donanemab. Over a 12-month cycle, blood is collected before initiation (BL, week 0), at treatment visits through week 16 and at week 52 (T26), until ARIA occurs or treatment is discontinued.

30nAdditional study group

AD patients not treated with amyloid immunotherapy

In some clinics approximately 90% of AD patients do not meet the criteria for amyloid immunotherapy; this group compares immune signatures of treated versus untreated patients.

30nControl

Patients ≥55 yrs · mild dementia, no AD

Aging persons with mild cognitive impairment who do not have AD (Figure 5).

30nControl

Patients ≥55 yrs · healthy

No dementia and no neurological diseases (Figure 5).

50nControl

Patients ≤55 yrs · healthy

Healthy younger participants, to understand changes in peripheral immune cells associated with disease onset (Figure 5).

Patient access through the Brain Health Program

The BHP provides access to patients treated with FDA-approved donanemab or lecanemab, who undergo MRI imaging, amyloid detection in cerebrospinal fluid and APOE genotyping. It has begun accepting referrals from general neurologists and geriatricians within the IU Health system.

102
patients enrolled who have begun therapy
39
additional patients at various stages of screening
As of November 5, 2024.
Interactive · Figure 6

Treatment protocol and blood-collection schedule

Treated patients take part in a 12-month treatment cycle. The proposal plans blood collection at the time of each treatment until ARIA occurs or treatment is discontinued; the research protocol samples only the time points drawn in Figure 6 — before initiation (BL, week 0), at treatment visits through week 16 and at week 52 (T26).

Select a time point.

MRI checkpoints · Lecanemab

  • M1MRI within 1 year prior to initiation · BL, week 0
  • M2MRI prior to 5th infusion · T5, week 10
  • M3MRI prior to 7th infusion · T7, week 14
  • M4MRI for selected patients · T26, week 52

MRI checkpoints as annotated in Figure 6 (infusions counted from T1, week 2); the MRI within 1 year prior to initiation is placed at the baseline visit.

Table view
Time pointWeekLecanemab (Leqembi)Donanemab (Kisunla)
BL0 blood draw · M1: MRI within 1 year prior to initiation blood draw · M1: MRI within 1 year prior to initiation
T12 Infusion #1 · blood draw Infusion #1 (700 mg) · blood draw
T24 Infusion #2 · blood draw No protocol visit
T36 Infusion #3 · blood draw Infusion #2 (700 mg) · blood draw · M2: MRI prior to 2nd infusion
T48 Infusion #4 · blood draw No protocol visit
T510 Infusion #5 · blood draw · M2: MRI prior to 5th infusion Infusion #3 (700 mg) · blood draw · M3: MRI prior to 3rd infusion
T612 Infusion #6 · blood draw No protocol visit
T714 Infusion #7 · blood draw · M3: MRI prior to 7th infusion Infusion #4 (1400 mg) · blood draw · M4: MRI prior to 4th infusion
T816 Infusion #8 · blood draw No protocol visit
T2652 Infusion #26 · blood draw · M4: MRI for selected patients blood draw · M5: MRI for selected patients

Adapted from proposal Figure 6. Study weeks count from the baseline visit (BL, week 0, before initiation); as drawn in the figure, therapy starts at T1 (week 2) and Tn is the n-th bi-weekly treatment, so T26 is week 52. Lecanemab is administered bi-weekly; donanemab monthly — 700 mg for the first three infusions, then 1400 mg, as stated in the proposal's Figure 6 (the current label uses a different titration). Blood draws: lecanemab BL, T1–T8, T26; donanemab BL, T1, T3, T5, T7, T26 — no protocol draws between week 16 and week 52. Lecanemab MRI checkpoints follow the Figure 6 annotations; donanemab checkpoints are the research portal's protocol. This is a research sampling schedule, not an ARIA-monitoring schedule for clinical use — see the current prescribing information for required MRI monitoring.

Aim 2 · Analysis

Compositional data and log-ratios

Flow cytometry data are typically compositional: each phenotype is a proportion of all phenotypes in a sample, with the total summing to 100%. If one population increases, another must decrease — so some phenotypes, or their absence at certain time points, can produce misleading correlations with the outcome.

The proposal addresses this with a temporal centered log-ratio: the proportion of each phenotype is divided by the geometric mean of that phenotype over all time points for the patient, and the logarithm of this value is reported, plotted and analyzed against the treatment timeline (Figure 6). This expresses each phenotype relative to the patient's own typical level.

On its own, the temporal version does not remove the sum constraint: if an abundant population such as neutrophils doubles at one visit, every other population's share falls at that visit and all of their temporal log-ratios drop together. The standard per-sample centered log-ratio (each population relative to the geometric mean of all populations in the same sample) removes the constraint; the research portal offers both transforms and labels them separately.

clrtemporal(pt,d)= log⁡pt,dGd , Gd=(∏t=1Tpt,d)1/T

pt,d: proportion of phenotype d at time point t; Gd: its geometric mean over the patient's T time points. A value of 0 is the patient's own typical level; +0.69 means twice that level, −0.69 half of it.

Worked example: one hypothetical patient

Hypothetical numbers

Proportion of leukocytes

On the percentage scale the rare population's halving is barely visible.

Temporal centered log-ratio (proposal)

Each population relative to its own geometric mean over time (dashed line = 0).
PopulationBL %T1 %T2 %T4 %T6 %T8 %Geometric meanFold change
Abundant population 55.0056.1054.2055.4056.0054.80 55.25×1.00
Intermediate population 6.006.206.907.808.408.90 7.29×1.48
Rare population 1.000.950.800.620.550.50 0.71×0.50

Temporal CLR values computed with the CogniMune analysis engine (temporal_clr), the same function used in the research portal. With only three illustrative populations the percentages do not sum to 100%; the example shows the temporal normalization, not closure.

How the analysis works

Optimal transport turns each immune profile into a graph fingerprint that can be compared over time — try the live Sinkhorn demo.

Technology Contact the team