Collide Therapeutics is building the first therapeutics against ZNF598 — the rate-limiting checkpoint of ribosome quality control, and the upstream failure point shared by ALS, Alzheimer’s and aging.
Thesis
Pharma has spent a decade failing to recognize these diseases as ribosomopathies.
Ribosomes are the most fundamental machines in biology. They execute all protein synthesis and consume the majority of a cell’s energy. When they stall — on damaged mRNA, rare codons, structured transcripts, poly(A) stretches — trailing ribosomes crash into the stalled one. That collision is the alarm.
Ribosome quality control is the system that answers it. ZNF598 is its most conserved sensor and its rate-limiting step. When ZNF598 is outpaced, unresolved stalls produce CAT-tailed proteins — among the most aggregation-prone species known — which seed TDP-43, SOD1 and tau pathology.
Every approved and failed program in ALS and Alzheimer’s sits either upstream of this checkpoint, silencing one mutation at a time, or downstream of it, clearing damage that has already happened. The checkpoint itself has never been drugged — because the structural basis of its key regulator eluded complete characterization for over a decade. Collide has resolved it.
“Increased ribosome pausing, leading to RQC overload and nascent polypeptide aggregation, critically contributes to proteostasis impairment and systemic decline during ageing.”
Ribosomes stall on damaged mRNA, rare codons and poly(A) stretches; trailing ribosomes collide. ZNF598 is the sensor that recognises the collided interface and marks it for resolution — and the rate-limiting step of the whole pathway. Mechanism per McGirr et al., FEBS J. 292(5), 2025 and De La Cruz et al., EMBO J. 44(18), 2025.
Competitive Landscape
Everything downstream has failed.
Map the assets in neurodegeneration onto the disease cascade and a shape appears: dense clusters at the genetic origin and at the downstream wreckage, and nothing at all on the quality-control checkpoint between them.
Gene silencers cover ≤15% of patients — each mutation separately. Downstream drugs have universally failed or shown marginal benefit. RQC is mutation-agnostic.
Gene
mutations
Gene silencing (ASOs)
Biogen – tofersen (SOD1)
Approved 2023 — 2% of patients
Biogen – BIIB078 (C9orf72)
Phase 1 discontinued
Ionis – ION363 (FUS)
Phase 1/3 — rare mutation
Biogen – BIIB105 (ATXN2)
Phase 1/2 discontinued (2024)
mRNA
transcription
eIF2B activators
Calico/AbbVie – fosigotifator
Ph 2/3 failed (Jan 2025)
Denali – DNL343
Ph 2/3 failed (Jan 2025)
Collide Therapeutics
First to drug RQC / ZNF598
Ribosome / RQC
(quality control)
Undrugged
Collide Therapeutics
First to drug RQC / ZNF598
Protein
aggregation
ER / mitochondrial stress
Amylyx – Relyvrio
Approved 2022 — withdrawn 2024
Mitsubishi – edaravone
Approved 2017 — mixed results
Motor neuron
stress
Glutamate reduction
Sanofi – riluzole
Approved 1995 — ~3mo survival
Degeneration
& paralysis
Supportive care only
No disease-modifying therapy
None available
UpstreamDownstream →
FailedApprovedCollide (undrugged)
Platform
Nine stages. Under a week.
A single automated run carries a target from an unsolved structure to a synthesizable lead: generate, dock, simulate, measure kinetics, optimize free energy, and plan the route. Every foundational tool in the stack is open-sourced, and the two models that do the work that nothing off-the-shelf could — RiBoltz2 and RiboFlow — are ours.
All foundational tools open-sourced
Runtime under a week
$500 – $2,500 per molecule
01RiBoltz2
Proprietary target structure
Resolves the ZNF598 binding modality that eluded complete structural characterization for over a decade.
02RiboFlow
Molecule generation
Generates novel chemical matter against the resolved site, conditioned on the target pocket.
03
Screen + pose
High-throughput docking and pose selection narrows generated libraries to credible binders.
04
MD stability
All-atom molecular dynamics tests whether a pose survives on a physical timescale.
05
Steered MD / PPI
Steered simulations probe the protein–protein interface and the force required to break it.
06WESTPA · SEEKR2
Kinetics
Weighted-ensemble and milestoning methods estimate residence time, not just affinity.
07
FEP optimization
Free-energy perturbation ranks analogues to drive the series toward potency.
08
Umbrella sampling
Free-energy landscapes along the binding coordinate confirm the mechanism of engagement.
09
Synthesis planning
Retrosynthetic routing converts a computational lead into something a chemist can make.
RiBoltz2
measured against published baselines
Memory efficiency — vs. a leading published structure-prediction baseline
250×Memory efficiencyvs. a leading published structure-prediction baseline
Memory efficiency — vs. a second published baseline model
8×Memory efficiencyvs. a second published baseline model
Accuracy — vs. a third published baseline model
6×Accuracyvs. a third published baseline model
State of the art for molecular glue generation and free-energy perturbation. Baseline models are unnamed pending publication; benchmark methodology is in preparation for submission to Nature Communications.
Pipeline
ZNF598 — a first-in-class target
One checkpoint, two directions. Enhancing ribosome quality control protects neurons that are losing the capacity to clear stalled translation; disrupting it removes the buffer that lets tumor cells tolerate translational stress. Both start from the same resolved structure.
ALS / C9orf72-FTD
Lead
RQC Enhancement
Lead program. Poly(GR) from the C9orf72 repeat expansion stalls ribosomes directly; ZNF598 titrates it.
Lead series
Alzheimer’s disease
RQC Enhancement
Inefficient stall resolution during APP synthesis generates CAT-tailed species that precipitate AD hallmarks.
Inverting the mechanism: removing the quality-control buffer makes translational stress lethal to tumor cells.
Exploratory
mRNA therapeutics optimization
Platform
Stall-aware sequence design to raise yield and lower immunogenicity of synthetic mRNA.
Exploratory
No disclosed clinical or preclinical program targets ZNF598. The lead series is moving out of silico: chemical synthesis and in vitro validation are the next milestone.
Research
Built on a decade of discovery
The link between ribosome quality control and neurodegeneration has been established across model systems and disease contexts — yeast, worm, killifish, fly and human cells; ALS, Alzheimer’s and aging. What this literature left structurally intractable was the druggability of the checkpoint itself. Collide’s models have resolved it, enabling rational design against the RQC pathway for the first time. That work is in preparation for submission to Nature Communications.
FEBS J.2025
Dysregulated ribosome quality control in human diseases
McGirr, T., Onar, O. & Jafarnejad, S.M. · 292(5):936–959