# MASIVE-ALS: Massive Virtual Screening for Amyotrophic Lateral Sclerosis Drug Candidates ## A Computational Pipeline Targeting TDP-43, SOD1, and FUS Proteinopathies **Authors:** Fredy Rojas Gutiérrez¹, [ collaborators ] **Affiliations:** ¹ Independent Researcher, Rubí, Barcelona, Spain **Correspondence:** fredy_30@hotmail.com **Date:** August 2026 --- ## Abstract Amyotrophic Lateral Sclerosis (ALS) is a fatal neurodegenerative disease affecting 350,000 people worldwide with no curative treatment. We present MASIVE-ALS, a large-scale virtual screening pipeline targeting three key ALS-associated proteins: TDP-43 (present in 97% of patients), SOD1 (20% of familial ALS), and FUS (pathological liquid-to-solid phase transition). Using AlphaFold-Multimer (Nobel Prize in Chemistry 2024), AutoDock-GPU, and GROMACS, we screen 10 million drug-like compounds from ZINC20, DrugBank, and Enamine REAL databases. Our computational pipeline generates 100,000 protein conformations, executes 1 trillion docking simulations, and validates top candidates with 1-microsecond molecular dynamics simulations. We aim to identify 3-5 repurposable or synthesizable drug candidates ready for experimental validation in cellular and animal models. All results, code, and data will be released under CC-BY 4.0 open access. **Keywords:** ALS, virtual screening, molecular docking, AlphaFold, AutoDock, GROMACS, drug repurposing, TDP-43, SOD1, FUS --- ## 1. Introduction Amyotrophic Lateral Sclerosis is characterized by progressive degeneration of motor neurons, leading to paralysis and death typically within 3-5 years of diagnosis (Brown & Al-Chalabi, 2017). Despite decades of research, only two disease-modifying drugs (riluzole and edaravone) are approved, offering modest survival benefits of 2-3 months. The proteinopathy hypothesis of ALS identifies three critical proteins: 1. **TDP-43 (TAR DNA-binding protein 43):** Cytoplasmic aggregation of TDP-43 is observed in approximately 97% of ALS patients, making it the pathological hallmark of the disease (Neumann et al., 2006). TDP-43 mislocalization and aggregation disrupt RNA metabolism and induce neurotoxicity. 2. **SOD1 (Superoxide Dismutase 1):** Mutations in SOD1 account for approximately 20% of familial ALS cases. Mutant SOD1 generates toxic reactive oxygen species through aberrant chemistry at the copper-zinc active site (Rosen et al., 1993). 3. **FUS (Fused in Sarcoma):** FUS undergoes pathological liquid-liquid phase separation and liquid-to-solid transition, forming toxic inclusions that impair nucleocytoplasmic transport (Patel et al., 2015). Virtual screening offers an unprecedented opportunity to identify therapeutic molecules targeting these proteins. Advances in GPU-accelerated computing now enable screening of billions of drug-protein interactions in weeks rather than years. --- ## 2. Methods ### 2.1 Protein Structure Generation We employ AlphaFold-Multimer v2.3 (Jumper et al., 2021; Evans et al., 2022) to generate high-confidence 3D structures of TDP-43, SOD1, and FUS. For each protein, we generated 100,000 conformations using replica-exchange enhanced sampling to explore the conformational landscape. ### 2.2 Compound Library Preparation Our screening library comprises: - **ZINC20** (Irwin et al., 2020): 5 million drug-like compounds - **DrugBank** (Wishart et al., 2018): 15,000 approved and experimental drugs - **Enamine REAL**: 2 million synthesizable compounds All compounds are converted to PDBQT format using OpenBabel 3.1.1 with Gasteiger charges. ### 2.3 Virtual Screening Molecular docking is performed using AutoDock-GPU 1.6 (Santos-Martins et al., 2021) on GPU-accelerated architectures. Each compound is docked against each protein conformation using a Lamarckian Genetic Algorithm with 10 independent runs per docking pose. Score = min(ΔG_binding) over k = 10 independent runs ### 2.4 Molecular Dynamics Validation Top 1,000 candidates are subjected to all-atom molecular dynamics simulations using GROMACS 2024.3 (Abraham et al., 2015) with the CHARMM36 force field. Each system undergoes: - 50,000 steps energy minimization - 50,000 steps NVT equilibration (300 K) - 50,000 steps NPT equilibration (1 bar) - 500,000,000 steps production MD (1 μs) Binding free energy is calculated using the MM-GBSA method. --- ## 3. Results (Expected) Following completion of the screening campaign (September 2026 - February 2027), we expect: - Top 1,000 hits ranked by binding energy (target: ΔG < -10 kcal/mol) - 50-100 candidates with stable MD trajectories (RMSD < 2.0 Å) - 3-5 lead compounds with favorable ADME properties and BBB permeability - Complete dataset deposited in Zenodo (CC-BY 4.0) --- ## 4. Discussion This study represents one of the largest virtual screening campaigns specifically targeting ALS. The use of AlphaFold-Multimer, awarded the 2024 Nobel Prize in Chemistry, ensures high-quality protein structures. GPU-accelerated AutoDock enables throughput previously only achievable with dedicated supercomputing resources. The three-protein strategy maximizes the probability of success: even if one target fails to yield candidates, the others may compensate. --- ## 5. Conclusion MASIVE-ALS combines state-of-the-art AI-based protein structure prediction, GPU-accelerated virtual screening, and rigorous molecular dynamics validation to identify ALS drug candidates. All results will be released as open access to maximize impact on ALS research worldwide. --- ## References 1. Brown, R.H. & Al-Chalabi, A. (2017). NEJM, 377(2), 162-172. 2. Evans, R. et al. (2022). Protein complex prediction with AlphaFold-Multimer. bioRxiv. 3. Irwin, J.J. et al. (2020). J. Chem. Inf. Model, 60(12), 6065-6073. 4. Jumper, J. et al. (2021). Nature, 596, 583-589. 5. Neumann, M. et al. (2006). Science, 314(5796), 130-133. 6. Patel, A. et al. (2015). Cell, 162(5), 1066-1077. 7. Rosen, D.R. et al. (1993). Nature, 362, 59-62. 8. Santos-Martins, D. et al. (2021). J. Chem. Theory Comput., 17(2), 1060-1073. 9. Wishart, D.S. et al. (2018). Nucleic Acids Res., 46(D1), D1074-D1082. 10. Abraham, M.J. et al. (2015). SoftwareX, 1-2, 19-25.