ASCEND-ALS: Advanced System for Clinical Evaluation and Novel Drug Design in ALS

Over 500 000 patients currently diagnosed with ALS have on average 2-5 years left to live with no cure or effective treatment plan to stop their muscles from becoming paralyzed. Time is crucial and every patient progresses differently. Care and treatment needs to be personalized.

To address this, we developed ASCEND-ALS, a two-branch computational platform:

ASCEND-Clinical is a complex machine learning system trained on over 30 000 patient visits. It diagnoses, predicts disease progression including major interventions such as feeding tube placement and ventilation, and all other clinical needs, enabling earlier and more tailored care.

ASCEND-Target which takes the gene of a lost neuronal protein essential for survival and generates ranked ASO candidates. It is a machine learning model trained on published experimental data, validated against three independent human datasets and wet lab experiments.

Our platform is unique: it spans diagnosis, disease progression, and therapy, aiming to accelerate drug development, reduce costs, and move ALS towards a truly personalized care.

Background and Inspiration

ALS is one of medicine's cruelest diseases destroying motor neurons in the brain and spinal cord and progressively affecting all muscles in the body while leaving the mind intact. Patients experience paralysis, difficulty speaking, swallowing and breathing with full cognitive awareness. Their diagnosis is frequently delayed, and it can take years for them to be screened by researchers for potential therapeutic treatments.

Our biggest inspiration was the patients we met as we researched ALS. One particular story shared was by a family doctor who told us about the moment he realized he had ALS. He had seen this disease in his own patients, understood his prognosis and yet nothing could tell him when, or what might help.

We learned about the challenges encountered by patients on their clinical journey, with ALS taking an average of 12 months to diagnose. Once diagnosed, patients face constant decisions about ventilation, nutrition, communication aids and palliative care. Decisions about these clinical milestones happen reactively without any predictive tools.

ALS researchers described to us the slow and expensive process of identifying drug candidates worth pursuing. Most potential therapeutics are screened experimentally with very little computational guidance.

These two groups became our “why”.

1. Clinical Interventions (ASCEND-Clinical)

A longitudinal machine learning framework was developed to model ALS progression using real patient data from three databases: Answer ALS, ALS TDI, and PROACT, spanning speech, swallowing, nutrition, mobility, and respiratory function, and over 10,000 patients.

Visit-level ALSFRS-R data were harmonized and transformed into engineered features: per-subscore slopes, domain trajectories, progression acceleration, baseline function, diagnostic delay, and King's ALS staging (Roche et al., 2012). A multi-model architecture was implemented across six stages: diagnosis, trajectory modelling, intervention forecasting, multidisciplinary care, advanced disease, and end of life:

Cox proportional hazards: time-to-event prediction for feeding tube, NIV/BiPAP, and tracheostomy

ElasticNet Cox: penalized survival regression with automatic feature selection

Random Survival Forest: nonlinear ensemble benchmark

DeepSurv: neural network connected to Cox partial likelihood

Models were validated using 5-fold cross-validation and cross-cohort external validation, benchmarked against ENCALS (Westeneng et al., 2018), performance measured by C-index and calibration.

2. Therapeutic Design (ASCEND-Target)

A modular computational pipeline was developed to design and prioritize antisense oligonucleotide (ASO) candidates targeting TDP-43-regulated cryptic exons, integrating genomic coordinates, biophysical scoring, transcriptome-wide safety screening, and patient-derived validation.

Cryptic exon coordinates for six genes were curated from published literature and validated by BLAT alignment against hg38. Candidate 20-mer ASOs were generated by sliding-window tiling and scored on an eight-feature composite model integrating:

RBP context

ViennaRNA accessibility

PhyloP conservation

TDP-43 binding affinity

SpliceAI delta scores

Sequence-level features

Off-target liability was screened by transcriptome-wide BLAST. A tiered Hill-curve dose-response framework was calibrated from Baughn 2023, with gene-specific protein-to-function models. Monte Carlo simulation generated confidence intervals; sensitivity analysis characterized dosing thresholds.

A gradient boosting model was trained on 1,488 labeled sequences including Ionis patent families (Tofersen, Spinraza). Validation spanned post-mortem ALS spinal cord, the Answer ALS iPSC cohort (n=1,276), and junction-level RNA-seq in the Kiskinis cell model.

1.Clinical Intervention Results (ASCEND-Clinical)

Result 1: All slope-augmented models exceeded the published ENCALS benchmark.

Five survival models were trained across patients with observed events using 5-fold cross-validation and cross-cohort external validation. Every slope-augmented model surpassed C-index ~0.78, the current gold standard for ALS prognostic modeling (Westeneng et al., 2018):

Cox baseline (11 features): C-index 0.776

Cox with slopes (46 features): C-index 0.864

ElasticNet Cox: C-index 0.865, external 0.873 / 0.828

Random Survival Forest: C-index 0.865, external 0.894 / 0.802

DeepSurv neural network: C-index 0.861

Result 2: Trajectory features, not model complexity, drove performance.

Adding rate-of-decline features contributed +0.075 C-index, 85% of total gain. Switching from Cox to a neural network contributed only +0.004. All four 46-feature models clustered within 0.005 C-index regardless of architecture, confirming ALS progression follows a largely linear hazard structure.

Result 3: ElasticNet identified a sparse, interpretable feature set.

L1/L2 regularization (Zou & Hastie, 2005) eliminated 7 of 46 features as redundant. Key surviving predictors:

Swallowing decline rate

Speech decline rate

King's staging, nutritional trajectory, and respiratory slope independently contributed across intervention forecasting and multidisciplinary care modules

Result 4: Isotonic recalibration achieved clinical-grade probability accuracy.

Brier score improved from 0.103 to 0.093 and Expected Calibration Error reached 0.000 making individualized risk predictions for feeding tube, ventilation, and palliative referral timing.

2.Therapeutic Design (ASCEND-Target)

Statistical Analysis:

Permutation tests (n=10,000) evaluated composite scoring against a null distribution. Pearson and Mann-Whitney U tests assessed target-gene dysregulation. Kolmogorov-Smirnov tests compared candidates against a random 20-mer baseline (n=1,000). Leave-one-sequence-out cross-validation evaluated ML generalization. Monte Carlo simulation (10,000 draws) generated 95% confidence intervals.

Results:

ASCEND reproduced the experimental potency rank-order of three reference ASOs (p<0.001). Baughn 2023 ASOs withheld from development were ranked in their exact measured EC50 order.

ASCEND identified an off-target liability missed by in vitro screening. rASO2 was excluded due to repeat-motif complementarity to five human transcripts — a risk BLAST detects but splice-reporter assays cannot.

STMN2 and UNC13A exhibit different protein-to-function architectures with dosing implications. STMN2 follows a linear model (Klim 2019, R²=0.94); UNC13A a sigmoid with a ~30% rescue threshold (Keuss 2024, R²=0.99). 69.4% of plausible parameter combinations predict the UNC13A threshold is crossed within the intrathecal therapeutic window.

Target genes show concordant dysregulation across ALS datasets. Five of six ASCEND targets were significantly downregulated in patient cervical spinal cord (Humphrey 2023, p<0.005). The largest iPSC motor neuron analysis to date (Answer ALS, n=1,276) confirmed STMN2 dysregulation at p=1.29×10⁻¹¹¹.

Junction-level RNA-seq validated both cryptic exon coordinates in patient iPSC motor neurons (Kiskinis GSE304550), with empirical decomposition of the splicing-to-protein ratio into NMD and intron-31 components.

Gradient boosting on 1,488 sequences (including Ionis patent families) achieved r=0.554 (p<0.0001), identifying sequence entropy, GC content, and MOE/PS backbone chemistry as dominant predictors.

1.ASCEND-Clinical

ASCEND-Clinical demonstrates that individualized ALS progression prediction is achievable. ALS has no cure. What it does have is time and right now, that time is lost because decline is recognized too late.

For patients, earlier prediction means earlier conversations about feeding, ventilation, and communication aids while they can still participate in those decisions. For families, it means time to prepare instead of responding to crises. For clinicians, it means a structured care plan built around each patient.

The platform also identifies patients likely to meet clinical trial eligibility windows, a critical bottleneck in ALS drug development where enrollment failure has derailed otherwise promising therapies.

Critically, this pipeline is the foundation for ASCEND-Clinical's broader goal: personalized care for ALS patients.

2.Therapeutic Design (ASCEND-Target)

ASCEND is the first computational ASO design platform to integrate patient-derived calibration with quantitative uncertainty propagation. Existing tools output point estimates; ASCEND reports Monte Carlo-derived 95% confidence intervals, transforming candidate ranking into a defensible framework for wet-lab prioritization.

The gene-architecture finding has direct clinical translatability: UNC13A’s sigmoid threshold versus STMN2’s linear response means dose regimens optimized for STMN2 will likely be sub-therapeutic for UNC13A — a mechanism-aware distinction absent from current ASO pipelines. Detection of an off-target liability in a peer-reviewed reference ASO further demonstrates that computational screening yields information complementary to wet-lab assays.

ALS has a 2-5 year median survival. ASCEND compresses 6-12 month wet-lab screening cycles into days of computation, ensuring candidates entering the lab are those most likely to advance.

The goal is to deploy ASCEND in ALS clinics and research labs as a unified decision-support platform — helping clinicians plan feeding, ventilation, and communication care proactively rather than reactively, while accelerating the path from target to validated drug candidate. The goal is to expand the platform with new patient cohorts and wet-lab data, extending predictions to additional milestones and disease genes. Wet-lab validation of top-ranked candidates follows. Above all, the goal is to personalize care and accelerate drug discovery — toward a world free of ALS.

This project would not exist without the people who believed in it before the results did.

David Taylor and ALS Canada have been our biggest supporters from day one. David opened doors we didn't know existed, and never once doubted us despite our inexperience. In a field this specialized, that kind of unconditional support is rare and it changed everything. ALS Canada as a whole gave us resources, community, and a stage.

To the professors and scientists who gave their time and expertise, thank you for taking two high school students seriously and for shaping how we think about this science.

To the ALS patients who shared their stories, this project exists because of you. Your experiences are why we started and why we kept going.

To our parents, our school, the BASEF community, and every delegate who challenged our thinking, thank you for making this possible.

[1] Alajangi, H. K., Kaur, M., Sharma, A., Rana, S., Thakur, S., Chatterjee, M., Singla, N., Jaiswal, P. K., Singh, G., & Barnwal, R. P. (2022). Blood–brain barrier: Emerging trends on transport models and new-age strategies for therapeutics intervention against neurological disorders. Molecular Brain, 15, 49. https://doi.org/10.1186/s13041-022-00937-4

[2] Alessandrini, F., Wright, M., & Cleveland, D. W. (2025). TDP-43 dysfunction compromises UPF1-dependent mRNA metabolism in ALS. Neuron. [Data set: GSE304550]. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE304550

[3] Answer ALS Foundation. (2023). Answer ALS multi-omics dataset [Multimodal data set]. Neuromine Data Portal. https://dataportal.answerals.org

[4] Atassi, N., Berry, J., Shui, A., Zach, N., Sherman, A., Sinani, E., Walker, J., Katsovskiy, I., Schoenfeld, D., Cudkowicz, M., & Leitner, M. (2014). The PRO-ACT database: Design, initial analyses, and predictive features. Neurology, 83(19), 1719–1725. https://doi.org/10.1212/WNL.0000000000000951

[5] Barker, S. J., Thayer, M. B., Kim, C., Tatarakis, D., Simon, M. J., Dial, L., Nilewski, R., Wells, C., Zhou, Y., Afetian, M., Akkapeddi, P., Chappell, A., Chew, K. S., Chow, J., Clemens, A., Discenza, C. B., Dugas, J. C., Dwyer, C., Earr, T., … DeVos, S. L. (2024). Targeting the transferrin receptor to transport antisense oligonucleotides across the mammalian blood-brain barrier. Science, 385(6706), eado1517. https://doi.org/10.1126/science.ado1517

[6] Baughn, M. W., Melamed, Z., López-Erauskin, J., Beccari, M. S., Ling, K., Zuberi, A., Presa, M., Gonzalo-Gil, E., Maimon, R., Vazquez-Sanchez, S., Chaturvedi, S., Bravo-Hernández, M., Taupin, V., Moore, S., Artates, J. W., Acks, E., Ndayambaje, I. S., Agra de Almeida Quadros, A. R., Jafar-Nejad, P., Rigo, F., Bennett, C. F., Lutz, C., Lagier-Tourenne, C., & Cleveland, D. W. (2023). Mechanism of STMN2 cryptic splice-polyadenylation and its correction for TDP-43 proteinopathies. Science, 379(6637), 1140–1149. https://doi.org/10.1126/science.abq5622

[7] Baxi, E. G., Thompson, T., Li, J., Kaye, J. A., Lim, R. G., Wu, J., Ramamoorthy, D., Lima, L., Vaibhav, V., Matlock, A., Frank, A., Coyne, A. N., Landin, B., Ornelas, L., Mosmiller, E., Thrower, S., Farr, S. M., Panther, L., Gomez, E., … Bhanu, B. (2022). Answer ALS, a large-scale resource for sporadic and familial ALS combining clinical and multi-omics data from induced pluripotent cell lines. Nature Neuroscience, 25(2), 226–237. https://doi.org/10.1038/s41593-021-01006-0

[8] Brown, A.-L., Wilkins, O. G., Keuss, M. J., Kargbo-Hill, S. E., Zanovello, M., Lee, W. C., Bampton, A., Lee, F. C. Y., Masino, L., Qi, Y. A., Bryce-Smith, S., Gatt, A., Hallegger, M., Fagegaltier, D., Phatnani, H., NYGC ALS Consortium, Newcombe, J., Gustavsson, E. K., Seddighi, S., Reyes, J. F., Ramos, D., Schiavo, G., Fisher, E. M. C., Raj, T., Secrier, M., Lashley, T., Ule, J., Buratti, E., Humphrey, J., Ward, M. E., & Fratta, P. (2022). TDP-43 loss and ALS-risk SNPs drive mis-splicing and depletion of UNC13A. Nature, 603(7899), 131–137. https://doi.org/10.1038/s41586-022-04436-3

[9] Calvo, A. C., Olivan, S., Manzano, R., Zaragoza, P., Aguilera, J., & Osta, R. (2012). Fragment C of tetanus toxin: New insights into its neuronal signaling pathway. International Journal of Molecular Sciences, 13(6), 6883–6901. https://doi.org/10.3390/ijms13066883

[10] Cedarbaum, J. M., Stambler, N., Malta, E., Fuller, C., Hilt, D., Thurmond, B., & Nakanishi, A. (1999). The ALSFRS-R: A revised ALS functional rating scale that incorporates assessments of respiratory function. Journal of the Neurological Sciences, 169(1–2), 13–21. https://doi.org/10.1016/S0022-510X(99)00210-5

[11] Cox, D. R. (1972). Regression models and life-tables. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 34(2), 187–202. https://doi.org/10.1111/j.2517-6161.1972.tb00899.x

[12] Ding, L., Kshirsagar, P., Agrawal, P., & Murry, D. J. (2025). Crossing the blood–brain barrier: Innovations in receptor- and transporter-mediated transcytosis strategies. Pharmaceutics, 17(6), 731. https://doi.org/10.3390/pharmaceutics17060731

[13] Giri, A., Shin, Y., Ha, J., Chae, S., Shrestha, P., Pham, K., Kwon, T. K., Park, J. H., Jeong, J., Lee, N. K., & Yook, S. (2025). Design of cathepsin-sensitive linkers for tumor-selective bioconjugate drug delivery. Journal of Controlled Release, 382, 106349. https://doi.org/10.1016/j.jconrel.2025.106349

[14] Goto, A., Yamamoto, S., & Iwasaki, S. (2022). Biodistribution and delivery of oligonucleotide therapeutics to the central nervous system: Advances, challenges, and future perspectives. Biopharmaceutics & Drug Disposition, 44(1), 26–47. https://doi.org/10.1002/bdd.2338

[15] Guo, C., Chen, K., Vatsavayai, S. C., Akiyama, T., Zeng, Y., Liu, C., Sianto, O., Yang, E., Bombosch, J., Powell, R., Zhen, S., Mekhoubad, S., Morrie, R. D., Miller, G., Green, E. M., Petrucelli, L., Seeley, W. W., & Gitler, A. D. (2025). Cryptic splicing in synaptic and membrane excitability genes links TDP-43 loss to neuronal dysfunction. bioRxiv. https://doi.org/10.1101/2025.08.28.672801

[16] Gupta, A., Ramanathan, R., Raulji, C. M., & Mahato, R. I. (2025). Nanoparticulate delivery and targeting of RNA to the brain. Biochimica et Biophysica Acta – Reviews on Cancer, 1881(1), 189232. https://doi.org/10.1016/j.bbcan.2025.189232

[17] Harrell, F. E., Califf, R. M., Pryor, D. B., Lee, K. L., & Rosati, R. A. (1982). Evaluating the yield of medical tests. JAMA, 247(18), 2543–2546. https://doi.org/10.1001/jama.1982.03320430047030

[18] Havens, M. A., & Hastings, M. L. (2016). Splice-switching antisense oligonucleotides as therapeutic drugs. Nucleic Acids Research, 44(14), 6549–6563. https://doi.org/10.1093/nar/gkw533

[19] Hsia, H., Zanini, C., Simonneau, C., Fraidling, J., Kraft, T. E., Mayer, K., Sommer, A., Indlekofer, A., Wirth, T., Benz, J., Georges, G., Langer, L. M., Gassner, C., Larraillet, V., Lohnmann, S., Koller, E., Manso, M., Ravn, J., Hofer, K., … Brinkmann, U. (2025). Improved targeted delivery of antisense oligonucleotide with an antibody mask. Nucleic Acids Research, 53(5), gkaf253. https://doi.org/10.1093/nar/gkaf253

[20] Humphrey, J., Venkatesh, S., Hasan, R., Herb, J. T., de Paiva Lopes, K., Küçükali, F., Byrska-Bishop, M., Evani, U. S., Narzisi, G., Fagegaltier, D., NYGC ALS Consortium, Sleegers, K., Phatnani, H., Knowles, D. A., Fratta, P., & Raj, T. (2023). Integrative transcriptomic analysis of the amyotrophic lateral sclerosis spinal cord implicates glial activation and suggests new risk genes. Nature Neuroscience, 26(1), 150–162. https://doi.org/10.1038/s41593-022-01205-3

[21] Hua, Y., Sahashi, K., Rigo, F., Hung, G., Horev, G., Bennett, C. F., & Krainer, A. R. (2010). Peripheral SMN restoration is essential for long-term rescue of a severe spinal muscular atrophy mouse model. Genes & Development, 24(13), 1487–1493. https://doi.org/10.1101/gad.1941110

[22] Ionis Pharmaceuticals. (2013). Compositions and methods for modulation of SMN2 splicing (U.S. Patent No. 8,361,977). U.S. Patent and Trademark Office. https://patents.google.com/patent/US8361977B2

[23] Ionis Pharmaceuticals. (2019). Compositions for modulating SOD-1 expression (U.S. Patent No. 10,385,341). U.S. Patent and Trademark Office. https://patents.google.com/patent/US10385341B2

[24] Ishwaran, H., Kogalur, U. B., Blackstone, E. H., & Lauer, M. S. (2008). Random survival forests. The Annals of Applied Statistics, 2(3), 841–860. https://doi.org/10.1214/08-AOAS169

[25] Jaganathan, K., Kyriazopoulou Panagiotopoulou, S., McRae, J. F., Darbandi, S. F., Knowles, D., Li, Y. I., Garg, A., Quintavalle, I., Levanon, E. Y., & Ye, C. (2019). Predicting splicing from primary sequence with deep learning. Cell, 176(3), 535–548. https://doi.org/10.1016/j.cell.2018.12.015

[26] Katzman, J. L., Shaham, U., Cloninger, A., Bates, J., Jiang, T., & Kluger, Y. (2018). DeepSurv: Personalized treatment recommender system using a Cox proportional hazards deep neural network. BMC Medical Research Methodology, 18(1), 24. https://doi.org/10.1186/s12874-018-0482-1

[27] Keuss, M. J., Miguez, R., Vandenberghe, R., Fratta, P., & Ward, M. E. (2024). Rescue of UNC13A expression by targeting cryptic exon splicing. bioRxiv. https://doi.org/10.1101/2024.06.20.599684

[28] Klim, J. R., Williams, L. A., Limone, F., Guerra San Juan, I., Davis-Dusenbery, B. N., Mordes, D. A., Burberry, A., Steinbaugh, M. J., Gamage, K. K., Kirchner, R., Moccia, R., Cassel, S. H., Chen, K., Wainger, B. J., Woolf, C. J., & Eggan, K. (2019). ALS-implicated protein TDP-43 sustains levels of STMN2, a mediator of motor neuron growth and repair. Nature Neuroscience, 22(2), 167–179. https://doi.org/10.1038/s41593-018-0300-4

[29] Krakora, D., Macrander, C., & Suzuki, M. (2012). Neuromuscular junction protection for the potential treatment of amyotrophic lateral sclerosis. Neurodegenerative Diseases, 10(1–4), 68–74. https://doi.org/10.1159/000335195

[30] Lorenz, R., Bernhart, S. H., Höner zu Siederdissen, C., Tafer, H., Flamm, C., Stadler, P. F., & Hofacker, I. L. (2011). ViennaRNA package 2.0. Algorithms for Molecular Biology, 6, 26. https://doi.org/10.1186/1748-7188-6-26

[31] Melamed, Z., López-Erauskin, J., Baughn, M. W., Zhang, O., Drenner, K., Sun, Y., Freyermuth, F., McMahon, M. A., Beccari, M. S., Artates, J. W., Ohkubo, T., Rodriguez, M., Lin, N., Wu, D., Bennett, C. F., Rigo, F., Da Cruz, S., Ravits, J., Lagier-Tourenne, C., & Cleveland, D. W. (2019). Premature polyadenylation-mediated loss of stathmin-2 is a hallmark of TDP-43-dependent neurodegeneration. Nature Neuroscience, 22(2), 180–190. https://doi.org/10.1038/s41593-018-0293-z

[32] Mhaske, A., Shukla, S., Ahirwar, K., Singh, K. K., & Shukla, R. (2024). Receptor-assisted nanotherapeutics for overcoming the blood–brain barrier. AAPS PharmSciTech, 25(4), 83. https://doi.org/10.1208/s12249-024-02802-w

[33] Oligonucleotide Therapeutics Society. (2021, June 21). Can oligonucleotides target RNA splicing to treat disease? https://oligotherapeutics.org/can-oligonucleotides-target-rna-splicing-to-treat-disease/

[34] Oligonucleotide Therapeutics Society. (2025, February 12). Transporting therapeutics: A novel approach to deliver ASOs to the brain. https://oligotherapeutics.org/transporting-therapeutics-a-novel-approach-to-deliver-asos-to-the-brain/

[35] Peng, X., Liu, X., Liu, Y., Kim, J. Y., Chen, Y., Ang, P., Nguyen, A., Leal, J., Yeh, H., & Ghosh, D. (2019). Brain-penetrating peptide shuttles across the blood-brain barrier and extracellular-like space. ChemRxiv. https://doi.org/10.26434/chemrxiv.8242871.v1

[36] Peterson, J. J., & Meares, C. F. (1998). Cathepsin substrates as cleavable peptide linkers in bioconjugates, selected from a fluorescence quench combinatorial library. Bioconjugate Chemistry, 9(5), 618–626. https://doi.org/10.1021/bc980039k

[37] Prades, R., Teixido, M., & Oller-Salvia, B. (2025). New trends in brain shuttle peptides. Pharmaceutics, 17(2), 238. https://doi.org/10.3390/pharmaceutics17020238

[38] Proudfoot, M., Jones, A., Talbot, K., Al-Chalabi, A., & Turner, M. R. (2016). The ALSFRS as an outcome measure in therapeutic trials and its relationship to symptom onset. Amyotrophic Lateral Sclerosis and Frontotemporal Degeneration, 17(5–6), 414–425. https://doi.org/10.3109/21678421.2016.1140786

[39] QurAlis Corporation. (2022). A study of QRL-201 in adults with ALS (NCT05633459). ClinicalTrials.gov. https://clinicaltrials.gov/study/NCT05633459

[40] Roche, J. C., Rojas-Garcia, R., Scott, K. M., Scotton, W., Ellis, C. E., Burman, R., Wijesekera, L., Turner, M. R., Leigh, P. N., Shaw, C. E., & Al-Chalabi, A. (2012). A proposed staging system for amyotrophic lateral sclerosis. Brain, 135(3), 847–852. https://doi.org/10.1093/brain/awr351

[41] Sanchez-Navarro, M., & Giralt, E. (2022). Peptide shuttles for blood–brain barrier drug delivery. Pharmaceutics, 14(9), 1943. https://doi.org/10.3390/pharmaceutics14091943

[42] Sethi, B., Kumar, V., Mahato, K., Coulter, D. W., & Mahato, R. I. (2022). Recent advances in drug delivery and targeting to the brain. Journal of Controlled Release, 350, 668–687. https://doi.org/10.1016/j.jconrel.2022.08.051

[43] Strom, A., Gal, J., Shi, P., Hayward, L. J., & Zhu, H. (2015). Retrograde axonal transport and motor neuron disease. Journal of Neurochemistry, 133(1), 8–21. https://doi.org/10.1111/jnc.13018

[44] Van Calster, B., Nieboer, D., Vergouwe, Y., De C, B., Pencina, M. J., & Steyerberg, E. W. (2016). A calibration hierarchy for risk models was defined: From utopia to empirical data. Journal of Clinical Epidemiology, 74, 167–176. https://doi.org/10.1016/j.jclinepi.2015.12.005

[45] Westeneng, H. J., Debray, T. P. A., Visser, A. E., van Eijk, R. P. A., Rooney, J. P. K., Calvo, A., Martin, S., McDermott, C. J., Dequen, P., Soufi, M., Beghi, E., Logroscino, G., Chiò, A., Hardiman, O., Braun, N., Al-Chalabi, A., & van den Berg, L. H. (2018). Prognosis for patients with amyotrophic lateral sclerosis: Development and validation of a personalised prediction model. The Lancet Neurology, 17(5), 423–433. https://doi.org/10.1016/S1474-4422(18)30089-9

[46] Zou, H., & Hastie, T. (2005). Regularization and variable selection via the elastic net. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 67(2), 301–320. https://doi.org/10.1111/j.1467-9868.2005.00503.x