AI Drug Discovery & Development for DNA Damage Response (DDR) Oncology Therapeutics
Submitted by Accelero Biostructures.
Originally published in Life Sciences Insights Magazine, August 2025
Irrespective of where one lies in the spectrum of Artificial Intelligence (AI) being real or too hyped, it is undeniable that massive growth in computing power and availability of massive amounts of data, has forever changed the AI field that now serves as an umbrella of a few sub-fields: data science/data analytics, machine learning/deep learning/neural networks and large language models. These results have deeply penetrated everyday lives, from casual online activities like shopping and movie recommendations to personal items like searching photo albums to offline activities like autonomous vehicles. Like technology companies, life sciences, drug discovery, drug development, medical science and healthcare companies have also joined the AI bandwagon.
In drug discovery and development, thanks to many years of development of public databases around biology, for example, genes and genomics; proteins and proteomics; target biology; and clinical trials; AI methods are being explored and applied for all components, including target discovery and validation, hit discovery, lead generation, preclinical studies, and clinical and regulatory studies. Our AI Drug Discovery/Design/Development (AIDD broadly) platform at Accelero Biostructures (Accelero) is founded on the use of protein-ligand interaction data from structural biology/co-crystallography on various drug targets, generated through our other platforms, in combination with other bioassay data on those targets. Much of the current AI on protein-ligand data involves the use of data from public databases, leading to AI models that lack diversity, uniqueness and negative data, all important components for training AI models. Our platforms generate high impact, target-specific, proprietary, positive and negative experimental data with good feedback loops that are critical for AI model training and optimization.
A decade (2015–2025) of California-based biopharma-directed high-throughput protein crystallography-based drug discovery at Accelero1 is founded on 15 years (2000–2015) of experience in structural genomics2, a field that was born following the human genome project and other genome sequencing projects3, and which we performed at California-based Berkeley Structural Genomics Center (BSGC)4 and the Joint Center for Structural Genomics (JCSG)5. Massive structural genomics efforts to experimentally determine novel protein crystal structures to populate the Protein Data Bank (PDB) ultimately led to the success of AlphaFold.6
Accelero’s AIDD focus is on early drug discovery and development to rapidly accelerate hit-to-lead generation and for maximizing clinical success. A critical part of this is to generate novel positive and negative experimental data that can be used for training and refinement of AI models to then use them for predictive optimization of chemical entities that can be quicker and more successful as they go through discovery and development. Our generation of novel data starts with proprietary experimental data on protein-ligand interactions from protein X-ray crystallography-based screening and co-crystallography of compounds bound to drug targets, which then feeds into partner drug discovery programs, for example, at XPose Therapeutics7, and Legend Innovation Life Science Fund8, that generate other streams of proprietary experimental data.
Specifically, as in the title of this article, this feeds into the AIDD program of one of our partners, XPose Therapeutics (XPose), focused on the discovery and development of novel inhibitors targeting DNA Damage Response (DDR) for oncology therapeutics. Cancer cells respond to increases in DNA damage, from elevated metabolism or exposure to therapeutic genotoxins, by upregulating their DDR. Several DDR pathways, each with multiple proteins, repair DNA damage in cancer cells, thereby prolonging cancer. Targeting DDR is a viable strategy in the treatment of cancer in both combinatorial therapies and monotherapies involving cancer-selective synthetic lethality. Despite immense potential of targeting DDRs, only a few targets are currently clinically targeted (PARP1) or near-clinically targeted (for example ATM, ATR, WRN), highlighting challenges in targeting DNA-binding DDR proteins (larger binding surfaces, charged), including nucleases, helicases, polymerases and transcription factors, proteins often refractory to drug discovery or “undruggable”. Our AIDD, coupled with other new and conventional drug discovery approaches, has the potential to propel DDR drug discovery to advance additional targets. XPose generates novel experimental data including new compounds; in vitro biochemical, biophysics, cell biology and ADME/PK assays; and in vivo data. Our data on diverse DDR targets, differentiated due to data novelty including protein-ligand structural data, will ultimately narrow the development path responding best to DDR inhibitors. We already see promising results with target APE1 in high-grade serous ovarian cancer.
References
1. Wilson DM 3rd, Deacon AM, Duncton MAJ, Pellicena P, Georgiadis MM, Yeh AP, Arvai AS, Moiani D, Tainer JA, Das D. Fragment- and structure-based drug discovery for developing therapeutic agents targeting the DNA Damage Response. Prog Biophys Mol Biol. 2021 Aug;163:130-142. doi: 10.1016/j.pbiomolbio.2020.10.005. Epub 2020 Oct 25. PMID: 33115610; PMCID: PMC8666131.
2. Zarembinski TI, Hung LW, Mueller-Dieckmann HJ, Kim KK, Yokota H, Kim R, Kim SH. Structure-based assignment of the biochemical function of a hypothetical protein: a test case of structural genomics. Proc Natl Acad Sci U S A. 1998 Dec 22;95(26):15189-93. doi: 10.1073/pnas.95.26.15189. PMID: 9860944; PMCID: PMC28018.
3. Burley SK, Almo SC, Bonanno JB, Capel M, Chance MR, Gaasterland T, Lin D, Sali A, Studier FW, Swaminathan S. Structural genomics: beyond the human genome project. Nat Genet. 1999 Oct;23(2):151-7. doi: 10.1038/13783. PMID: 10508510.
4. Shin DH, Hou J, Chandonia JM, Das D, Choi IG, Kim R, Kim SH. Structure-based inference of molecular functions of proteins of unknown function from Berkeley Structural Genomics Center. J Struct Funct Genomics. 2007 Sep;8(2-3):99-105. doi: 10.1007/s10969-007-9025-4. Epub 2007 Sep 2. PMID: 17764033.
5. Elsliger MA, Deacon AM, Godzik A, Lesley SA, Wooley J, Wüthrich K, Wilson IA. The JCSG high-throughput structural biology pipeline. Acta Crystallogr Sect F Struct Biol Cryst Commun. 2010 Oct 1;66(Pt 10):1137-42. doi: 10.1107/S1744309110038212. Epub 2010 Sep 30. PMID: 20944202; PMCID: PMC2954196.
6. Callaway E. The huge protein database that spawned AlphaFold and biology’s AI revolution. Nature. 2024 Oct;634(8036):1028-1029. doi: 10.1038/d41586-024-03423-0. PMID: 39424937.
Get the latest thought leadership from California’s life sciences sector in the quarterly Life Sciences Insights magazine. Share your own news and insights with CLS Member Voice.
FAQ: AI Drug Discovery for DDR Oncology Therapeutics
AIDD applies artificial intelligence across the drug discovery and development pipeline — including target discovery and validation, hit discovery, lead generation, and preclinical and clinical studies. Accelero Biostructures’ AIDD platform is built on proprietary protein-ligand interaction data from structural biology and co-crystallography, combined with bioassay data on specific drug targets.
Many AI models rely on public databases, which often lack diversity, uniqueness, and negative data — all essential for robust training. Accelero generates high-impact, target-specific, proprietary positive and negative experimental data with strong feedback loops, which are critical for training and optimizing predictive models.
DDR is the set of pathways cancer cells use to repair DNA damage — for example, from elevated metabolism or therapeutic genotoxins — which helps prolong the cancer. Targeting DDR is a viable strategy in both combination therapies and monotherapies that exploit cancer-selective synthetic lethality.
Many DDR proteins bind DNA and have larger, charged binding surfaces — including nucleases, helicases, polymerases, and transcription factors — which makes them often refractory to conventional drug discovery, or “undruggable.” Only a few targets are currently clinically targeted (such as PARP1) or near-clinically targeted (for example, ATM, ATR, and WRN).
Accelero generates proprietary protein-ligand structural data through X-ray crystallography-based screening and co-crystallography, which feeds partner programs such as XPose Therapeutics’ DDR effort. Coupled with conventional approaches, this novel data helps advance additional DDR targets — with promising early results reported for target APE1 in high-grade serous ovarian cancer.