Amyotrophic lateral sclerosis is a hard target-discovery problem. The disease is heterogeneous in its genetics, in its anatomical onset, and in its clinical course. The four headline causal genes — C9orf72, SOD1, FUS, and TARDBP — between them account for somewhere around 5–10% of all cases as familial disease, and may contribute in additional cases without a recognised family history (Meyer 2021). The remaining majority of patients have sporadic ALS without an identified Mendelian cause. Five FDA-approved drugs sit on the treatment side, each with modest benefit (Arnold et al. 2024), and the most recent — tofersen, an antisense oligonucleotide that reduces SOD1 expression — applies only to the small fraction of patients with SOD1 mutations (Everett & Bucelli 2024).
For target-discovery work, this is the question: how does a research team make sense of evidence that sits across multiple genetic causes and multiple cell types, all of which converge on motor-neuron loss but none of which is the same mechanism?
The textbook framing of ALS is “motor neuron disease,” and at a clinical level that is correct: the neurons that are lost are the upper and lower motor neurons. But for target-discovery purposes, treating ALS as a motor-neuron-only problem ignores most of the disease biology.
Glutamate-mediated excitotoxicity sits at the centre of the cellular cascade. Reviews of ALS pathogenesis describe a cortical-hyperexcitability model in which calcium-permeable AMPA receptors on motor neurons, dysfunction of the EAAT2 astrocytic glutamate transporter, increased presynaptic glutamate release, and reduced cortical-interneuron inhibition together drive a chronic excitotoxic load on the motor neurons that eventually kills them (Arnold et al. 2024). Three of those four mechanisms involve cells other than the motor neuron itself: astrocytes, presynaptic terminals (which may belong to other neurons), and interneurons.
The cellular consequences of that excitotoxic load are also non-cell-autonomous. Mitochondrial dysfunction, reactive-oxygen-species production, and endoplasmic-reticulum stress in the motor neurons are downstream of upstream activity in the surrounding glia and synapses (Arnold et al. 2024).
A target-discovery query against ALS that returns motor-neuron-expressed genes only is missing the cell types where the upstream biology lives.
Inside the Euretos AI Platform, the ALS view begins with the genetic architecture. The four established Mendelian genes — C9orf72, SOD1, FUS, TARDBP — anchor the integrated knowledge graph at the disease level, with each gene linking to its known protein-aggregation pathway, its model-system perturbation evidence, and the broader genetic-association literature. TDP-43 protein aggregates appear as a molecular hallmark across most ALS subtypes irrespective of which gene is causal (Meyer 2021).
The platform’s cell-type filter then makes the next analytical step concrete. Researchers can resolve target-discovery results for ALS to:
A target-discovery query restricted to each of these cell types returns a different ranked candidate list, integrated with the same genetic, perturbation, and literature evidence the platform applies elsewhere. The list against motor neurons is dominated by neuronal stress-response and protein-handling genes. The list against astrocytes is dominated by glutamate-handling and reactive-astrocyte programme genes. The list against microglia is dominated by neuroinflammatory and complement-pathway genes.
ALS is also a disease where Indication Selection — the platform’s other principal capability — meets target-discovery work in a useful way. Tofersen demonstrated that a genetically-defined ALS subtype (SOD1 mutation carriers) is a tractable target population for an antisense-oligonucleotide approach (Everett & Bucelli 2024). C9orf72 hexanucleotide-repeat expansions present a different but analogous opportunity. The platform’s Indication Selection view lets a researcher take a gene of interest from a different starting point — a candidate emerging from a sporadic-ALS perturbation experiment, for example — and ask which patient subgroups it might define.
The honest framing here is that ALS is harder than psoriasis. The biology runs across more cell types, the genetic architecture is more fragmented, and the disease modifying treatments to date have produced single-digit-percentage improvements rather than the step-change responses seen in some other indications. What an integrated platform can do is reduce the integration cost of the question. It cannot make the underlying biology any less hard.
The next case study in this series will look at indication selection more directly, using a worked example from a different therapeutic area.
Sources cited based on articles retrieved from PubMed.
Arnold et al. 2024 — Revisiting Glutamate Excitotoxicity in ALS, International Journal of Molecular Sciences
Meyer 2021 — Amyotrophic lateral sclerosis: diagnosis, course of disease and treatment options, Deutsche Medizinische Wochenschrift
Everett & Bucelli 2024 — Tofersen for SOD1 ALS, Neurodegenerative Disease Management