The molecular track: what held up, what failed, and why I pivoted.

I’m a modeler by trade: simulation, calibration, being honest about uncertainty. This is me pointing those tools at Alzheimer’s and ALS — the two diseases that are mine. The full personal story is here →

What I set out to do: start with protein aggregation — a mechanism implicated in both diseases — using public deep-mutational-scanning data where scientists have already measured what mutations do. That measurement is an answer key. If a model is wrong, the data says so immediately. The experiment already exists; no new wet-lab work is required to test the model against it. That built-in check is why a solo effort can make a checkable contribution instead of just writing reports.

The build

Step 1: Reproduce the field’s model

Before building anything I reproduced CANYA — a leading published protein aggregation model — on its own held-out data. AUROC 0.82 on 7,626 random peptides (the paper reports 0.809). Good: my pipeline is faithful, and I have a concrete baseline to improve on.

Step 2: Found and fixed its overconfidence

A model can rank things correctly and still lie about how sure it is. CANYA does. Its raw calibration error was 0.068 — systematically overconfident. Isotonic recalibration cut it ~6x, to 0.011, with its ranking completely unchanged. This is a standard technique, not a breakthrough — but it makes the model’s uncertainty more reliable. The thing I most want this mission to be known for is never handing someone a confidently wrong number about their disease.

Step 3: Built and tested an Aβ-specific disease model

Alzheimer’s has a handful of inherited mutations — dominantly-passed, causing early-onset familial AD — that make amyloid-β aggregate. I trained a from-scratch Aβ-specific model on 14,015 double-mutants and tested it on 468 held-out single-mutant outcomes, including all 8 familial outcomes. Their substitution features were represented among the double-mutant training examples.

Against the lab measurements, my model’s Spearman correlation was 0.62; zero-shot CANYA managed 0.20. On telling the familial mutations apart from the rest, my model reached AUROC 0.92 — matching the measured assay’s own discrimination ceiling (~0.90) and well above zero-shot CANYA (0.76). To be clear about provenance: that the assay separates the known familial mutations from the rest is the source lab’s own published finding (Seuma et al., eLife 2021), reproduced here — not a discovery of this work. What I add is a from-scratch model that recovers that separation and the honestly calibrated uncertainty around it.

Step 4: Tried to cross the protein boundary — and learned why it fails

I pointed the Aβ model at the ALS protein TDP-43. It didn’t just fail — it failed backwards (transfer Spearman -0.19). A hydrophobic substitution correlates with harm in the Aβ nucleation assay (+0.30) but with protection in the TDP-43 toxicity assay (-0.45). A stronger tested representation did not rescue the transfer. Because the assays measure different phenotypes, this result cannot separate protein biology from assay mismatch or rule out every representation. The comparison uses 1,196 TDP-43 measurements.

The reckoning

After the initial results, I kept going — twelve investigations in total. I tested two ESM-2 scales and an asymmetric tail-aware loss against the remaining weakness: systematic under-prediction of the most extreme-nucleating variants. None resolved that error while preserving the other gains. That is a negative result about the tested escalations, not proof that every model-side approach must fail. Structure-aware inputs and greater extreme-tail coverage are the next hypotheses to test.

I also tried to ask the boldest disease-framed question the data could answer: do second-site suppressors exist for the familial mutations? The answer is PARTIAL at its thinnest: one fragile lead, resting on two discordant replicates, losing significance on either alone — not a map, not a therapeutic claim (you cannot add a second mutation to a person).

Here is the honest account: the Lehner / Bolognesi lab has this comprehensively. Arutyunyan and colleagues published the 2025 energetic and thermodynamic model of Aβ aggregation with hundreds of measured couplings and a structural model of the transition state. Thompson and colleagues published CANYA, and the earlier Seuma et al. deep-mutational-scan paper established that the familial mutations are separable by their measured aggregation effect — the result I reproduce above, not one I discovered. My contribution on the molecular side is real — a rigorously reproduced, calibrated model with explicit uncertainty intervals, and a controlled result showing that the tested ESM-2 representation did not rescue cross-assay transfer — but it is a contribution built on their foundation, not a competing discovery.

The most defensible thing I built on the molecular track is not a new result about Aβ. It is a model that refuses to be confidently wrong about what it knows and doesn’t know — one whose prediction intervals are sharp where it is informed and honestly wide where it is blind. That’s the edge I came here to demonstrate. And it pointed toward what I should actually be spending time on.

The pivot

I started this mission to build the thing I most wanted to exist — honest foresight about disease progression, with real uncertainty bands, while there’s still time for it to matter.

At the time this chapter was published, the next problem looked like calibrated individual prognosis for ALS, starting with speech. Later access and feasibility decisions changed the sequence. The durable conclusion is narrower: the molecular work established the uncertainty and validation discipline that later prognosis studies would need.

For the formal results, intervals, sources, and complete limitations, read the molecular study → Current program status is maintained on Research. — Michael