Anger. For me, anger is the most authentic and motivating emotion ever. If I have a disagreement, my brain will churn sassy facts 100x faster than ever. I might not say a thing back to the person — but baby, I have crushed you 100 times in my mind. And the kick I get out of it? Pure case study.
My anger makes me driven. After every emotional spike, I refine my thinking, perspectives, and behavior. Most people struggle to change themselves. I actively audit myself.
For me, the most valuable thing is my energy and mindset.
If you stop growing while being around me, I will mentally relocate you to another planet — even if we share the same physical space.
Sapiogirl energy, I guess.
I am also an Enneagram 7. Having 100 interests is the bare minimum for me.
And lately?
One obsession: genomics, cell atlases, and age-related macular degeneration (AMD).
Because the more I read bioscience research, the more I realize something radical:
Science is not uncertain because it is weak.
Science is uncertain because biology is multidimensional.
The Illusion of “Not Knowing” in Medicine
In traditional medicine, diseases are treated like static labels:
“Wet AMD”
“Dry AMD”
“Intermediate stage”
“Responder vs non-responder”
But genomics shattered that illusion.
A Cell Atlas approach — mapping every cell type, state, and molecular pathway — reveals that a disease is not a single entity.
It is a dynamic ecosystem of cellular behaviors.
Instead of asking:
“What is AMD?”
Modern genomics asks:
“Which cells are malfunctioning, in which pathways, at which timepoint, in which patient subtype?”
That shift alone is the foundation of precision medicine.
What the Cell Atlas Really Means for Drug Development
The Human Cell Atlas and genomics-driven atlases (like those discussed by Illumina in drug development frameworks) are transforming how we understand disease at the cellular resolution.
Instead of bulk tissue averages, we now study:
Single-cell RNA expression, Cell-type specific pathways, Spatial transcriptomics, Disease-state cellular transitions
Why this matters for drug development:
- You can identify which cell populations actually drive disease
- You avoid treating the wrong biological target
- You design therapies tailored to molecular subtypes
In complex diseases like AMD, this is revolutionary.
Because AMD is NOT just a retinal disease.
It is: Genetic, Metabolic, Inflammatory, Neurovascular, Age-driven, Environmentally modulated
All at once.
Age-Related Macular Degeneration: A Precision Medicine Case Study
AMD is the leading cause of vision loss in the developed world, particularly in its neovascular form where abnormal blood vessels damage retinal structure and function.
But here’s the deeper truth: AMD progression is not linear. It is probabilistic and heterogeneous.
Two patients with the same ICD label can have:
- Completely different genomic risk
- Different RPE degeneration pathways
- Different inflammatory signatures
- Different conversion timelines to wet AMD
And yet historically, they receive the same treatment paradigm.
That is scientifically inefficient.
Genomics + Cell Atlas + AMD = The Future of Ophthalmology
Let’s connect the dots scientifically.
Retinal Cell-Level Complexity
The retina contains:
- Photoreceptors
- Retinal pigment epithelium (RPE)
- Microglia
- Choroidal vasculature
- Müller glia
A cell atlas approach helps us understand which of these cellular systems fail first in different AMD subtypes.
For example:
RPE dysfunction → geographic atrophy
Angiogenic signaling → neovascular AMD
Complement dysregulation → inflammatory progression
This aligns directly with genomic stratification efforts in precision ophthalmology.
How AI Actually Uses Cell Atlases for Drug Discovery (Not Sci-Fi, Real Science)
Now let’s move from philosophy to mechanism.
Because once you have a Cell Atlas, the real question becomes:
What do we do with this biological resolution?
This is where AI stops being buzzword and starts becoming a research instrument.
Not “AI will cure disease.”
But:
AI will help us ask better biological questions at the cellular level.
Especially in complex diseases like Age-related Macular Degeneration (AMD), where progression is not binary but deeply heterogeneous.
The Problem: Traditional Drug Discovery Ignores Cellular Heterogeneity
Most legacy drug discovery pipelines assume:
- One disease
- One pathway
- One target
- One drug
But a retina cell atlas shows something radically different:
- Stressed RPE cells
- Activated microglia
- Angiogenic endothelial cells
- Degenerating photoreceptors
All co-existing within the same “AMD” label.
So the real scientific question becomes:
Which cell states are actually driving disease progression to wet AMD?
Not just which gene is differentially expressed in bulk tissue.
A Hybrid AI Framework: Cell Atlas → Target Discovery → Drug Repurposing (AMD Example)
Let’s walk through a realistic AI pipeline used in modern precision drug discovery.
Step 1 — Learn Cellular States from the Retina Cell Atlas
Input:
- Single-cell RNA-seq retina atlas (healthy vs early AMD vs intermediate vs neovascular)
- Optional spatial transcriptomics
- Imaging biomarkers (OCT features)
AI Model:
- Variational Autoencoder (VAE) or scVI-style embedding
What it does:
- Compresses high-dimensional gene expression into latent biological representations
- Separates true biology from batch noise
- Identifies disease-specific cell states (e.g., “angiogenic endothelium” or “stressed RPE”)
This is critical for AMD because RPE dysfunction and angiogenesis are central to progression.
Step 2 — Link Cell States to Clinical Progression (Precision Medicine Layer)
Now we integrate:
- Cell state proportions
- Longitudinal EHR phenotypes
- Conversion to neovascular AMD
- Follow-up timelines
(Exactly the type of longitudinal data modern ophthalmology cohorts and EHR repositories capture.)
AI Model:
Gradient boosting / survival models / multimodal neural networks
Output:
Which cell states predict rapid progression
Which cellular pathways are high-risk signatures
Instead of saying:
“Patient has intermediate AMD”
We can say:
“Patient shows high angiogenic endothelial signature + inflammatory microglial activation = high conversion risk.”
That is true precision medicine.
Step 3 — AI Identifies Target Genes Within Disease-Critical Cell Types
Once high-risk cell states are identified, AI ranks:
- Differentially expressed genes
- Network-central regulators
- Genetically supported targets (GWAS/eQTL integration)
For AMD, this often highlights:
- VEGF pathway regulators
- Complement system genes
- RPE stress-response pathways
- Angiogenic signaling drivers
This is far more biologically grounded than bulk transcriptomics.
Step 4 — Drug Discovery via Signature Reversal (Where AI Becomes Powerful)
Now comes the most elegant step.
We construct a disease signature:
- Genes up-regulated in angiogenic endothelial cells
- Genes down-regulated in protective RPE states
Then AI screens compounds by asking:
Which drugs reverse this cellular disease signature?
Methods include:
Connectivity mapping (drug vs gene-expression reversal)
Graph Neural Networks on drug–target–pathway graphs
Multimodal models combining chemical structure + genomic targets + cell-state embeddings
This enables:
Novel drug discovery
OR intelligent drug repurposing
A Real-World Translational Signal: Dopamine Pathways in AMD
Interestingly, emerging longitudinal research has shown that L-DOPA exposure is associated with a reduced likelihood of conversion to neovascular AMD, suggesting that systemic therapies can influence retinal cellular pathways involved in angiogenesis.
From a cell-atlas perspective, this is biologically interpretable: retinal pigment epithelium (RPE) cells express receptors involved in angiogenic regulation, and modulation of these pathways can alter VEGF signaling and disease progression.
This reinforces a key insight — drugs do not act on disease labels; they act on cellular mechanisms.
Pseudocode: Conceptual AI Pipeline for Cell-Atlas–Driven Drug Discovery
Conceptually, this entire AI-driven drug discovery workflow can be summarized as a cell-atlas–guided computational pipeline rather than a traditional one-target screening approach:
# Input: Retina cell atlas + clinical progression data
sc_data = load_single_cell_atlas()
ehr_data = load_longitudinal_amd_cohort()
# Step 1: Learn biological embeddings
latent_space = train_scVI(sc_data)
# Step 2: Identify disease cell states
cell_states = cluster_cells(latent_space)
# Step 3: Link cell states to progression risk
features = aggregate_cell_state_scores(cell_states, ehr_data)
risk_model = train_survival_model(features, ehr_data["conversion_to_wet_amd"])
# Step 4: Extract key pathways
important_states = explain_model(risk_model)
target_genes = rank_genes(important_states)
# Step 5: Drug discovery via signature reversal
drug_candidates = screen_drugs_by_signature_reversal(target_genes)
return prioritized_drugs(drug_candidates)
Why This Matters Specifically for AMD (Not Just General AI Hype)
AMD is:
- Age-dependent
- Genetically influenced
- Environmentally modulated
- Clinically heterogeneous
Two patients with identical OCT and ICD labels can have completely different molecular trajectories.
A cell-atlas–guided AI framework allows:
- Subtype-specific therapies
- Earlier intervention before wet conversion
- Personalized treatment intervals
- Smarter clinical trial stratification
This is especially aligned with next-generation biomedical data platforms that integrate imaging, genomics, and longitudinal phenotyping.
Why Precision Medicine in AMD Needs Genomic Context (Not Just Imaging)
In clinical pipelines (including large phenotyping systems like the ones I work on), we often rely on:
- OCT biomarkers
- ICD codes
- Injection history
- Follow-up timelines
But genomics adds a missing layer:
- PRS (Polygenic Risk Scores)
- Complement pathway genetics (CFH, ARMS2)
- Cellular transcriptomic signatures
- Drug response heterogeneity
Imagine combining:
- EHR + Imaging + Genomics + Cell Atlas
Instead of treating AMD as a binary disease.
That is true next-generation data platforms in biomedicine.
The Future: Cell Atlases Will Redefine Drug Development in AMD
What excites me most is this:
Future AMD clinical trials will likely stratify patients based on:
- Genomic risk profiles
- Cell-type dysfunction signatures
- Molecular pathway activation
- Real-world longitudinal data
Not just visual acuity scores.
And that shift will:
- Reduce trial failures
- Enable targeted therapeutics
- Personalize treatment intervals
- Identify preventive therapies before wet conversion
And honestly?
That is the most exciting phase of science to live in — especially if you are working at the intersection of genomics, AI, and retinal disease like I am.