The Resistance Landscape: Four Disease Domains, One Systemic Failure

Extreme drug resistance is not a single biological event but a convergent failure mode appearing across oncology, HIV/AIDS, tuberculosis, and malaria — each driven by distinct but mechanistically overlapping processes. Patent filings and academic literature in this dataset map resistance mechanisms including target-site mutations, efflux pump overexpression, synthetic rescue pathway activation, and epigenetic reprogramming as the dominant modes of treatment failure.

In oncology, the EGFR T790M gatekeeper mutation — and its compound forms including the T790M/C797S double mutant and the exon19del/T790M/C797S triple mutant — is the most densely represented resistance target across retrieved results, appearing in at least eight papers spanning mechanism, structural drug design, and clinical strategy. Approximately 50% of acquired resistance to first- and second-generation EGFR tyrosine kinase inhibitors (TKIs) is attributable to the T790M substitution. The C797S mutation then confers resistance to third-generation agents such as osimertinib, creating a compound mutant configuration that is currently inaccessible to all three approved EGFR TKI generations.

Additional oncology resistance targets catalogued in this dataset include MET amplification, ERBB2 (HER2) amplification, ALK kinase domain mutations, BCR-ABL1 in CML, NTRK1, KDR, TGFBR2, PTPN11, CDK4, and CDKN2B — the latter identified by next-generation sequencing in patients with intrinsic TKI resistance. In HIV, resistance mutations span HIV-1 protease (with eleven darunavir resistance-associated mutations characterised), reverse transcriptase, and integrase. In TB and malaria, resistance emerges at the level of metabolic pathway networks and parasite kelch13 mutations respectively.

What unites these domains is the inadequacy of single-target, single-agent approaches in the face of tumour or pathogen plasticity. Resistance to one agent frequently accelerates the selection of cross-resistant clones, and the sequential deployment of improved monotherapies — the dominant commercial strategy for the past two decades — has proven insufficient. The field is now converging on three structural solutions: AI-driven target discovery, dual-pharmacophore hybrid molecules, and computationally optimised combination regimens.

Hybrid Molecules: Dual-Pharmacophore Design as a Resistance Bypass

Hybrid molecule design — the covalent linkage of two pharmacophore units with independent biological activity into a single agent — is the most extensively represented chemical modality for overcoming extreme resistance in this dataset, spanning oncology, HIV, and malaria. The approach, termed “covalent bitherapy” or “molecular hybridisation,” simultaneously addresses multiple resistance mechanisms and reduces the probability of complete cross-resistance compared with sequential monotherapy.

The breadth of validated hybrid scaffolds in the literature is notable. In malaria, dihydroartemisinin–HDAC inhibitor hybrids (compound α-7c) demonstrated single-digit nanomolar IC₅₀ against both artemisinin-sensitive (3D7) and artemisinin-resistant (Dd2) Plasmodium falciparum strains, with concurrent activity against leukemia cell lines — a dual-indication anti-resistance scaffold published by Heinrich-Heine University Düsseldorf in 2022. In oncology, conjugation of TKIs to platinum-based drugs produced agents that retained oncogenic kinase specificity while escaping efflux transporter-mediated resistance, representing the first report applying hybrid drug design to convert TKIs from P-glycoprotein substrates to non-substrates — with implications for CNS metastasis penetration, from the Chinese University of Hong Kong.

In HIV, the zidovudine (AZT) pharmacophore has been deployed as a scaffold for molecular hybridisation targeting HIV integrase, reverse transcriptase, and protease simultaneously — addressing resistance to combined antiretroviral therapy (cART). A systematic review from the University of Cadi Ayyad documented hybrid structures targeting all three major HIV enzymatic targets, emphasising reduced toxicity relative to parent pharmacophores and improved activity against resistant strains. According to WHO, multi-drug resistant HIV remains a critical global health challenge, making multi-target hybrid approaches particularly strategically relevant.

“Hybrid molecule design consistently outperforms monotherapy across oncology, HIV, and malaria — converting P-glycoprotein substrates to non-substrates, achieving single-digit nanomolar activity against resistant strains, and simultaneously engaging integrase, reverse transcriptase, and protease in a single chemical entity.”

From an IP strategy perspective, the retrieved dataset reveals that commercial patent activity in hybrid molecule design for extreme resistance is sparse. The dominant innovation signal comes from academic literature, suggesting that IP space in this area remains largely unoccupied by commercial filers — a structural opportunity for drug developers willing to translate academic scaffolds into proprietary chemical series.

AI and Computational Target Discovery: From TCGA to Exascale

AI and machine learning approaches to drug resistance represent the fastest-growing methodological cluster in this dataset, spanning genome-wide synthetic rescue prediction, parabolic response surface platforms for TB, crowdsourced polypharmacology challenges, and patient-specific exascale precision medicine frameworks. These approaches identify resistance-mediating targets that are structurally invisible to conventional biochemical target identification.

Synthetic Rescue Gene Prediction

The most mechanistically novel AI target class in this dataset is the synthetic rescue (SR) rescuer gene — computationally predicted from transcriptomics and survival data across 10,000 TCGA cancer patients by teams at UC Berkeley and Tel-Aviv University. SR rescuer genes are adaptive alterations that compensate for drug-induced target inactivation; inhibiting the predicted rescuer synergistically sensitised resistant cells in experimental validation. This target class is not accessible by traditional target identification methods, and commercial IP filings in this space appear absent from the retrieved dataset — a significant unoccupied IP opportunity.

Machine Learning for HIV Resistance Landscapes

At Institut Pasteur and CNRS, random forest and multi-label classifiers trained on 662 protease sequences and 715 reverse transcriptase sequences predicted cross-resistance patterns across antiretroviral regimens — enabling pre-treatment identification of regimen failure risks. Separately, machine learning analyses of approximately 55,000 HIV-1 RT sequences identified novel epistatic interactions between resistance mutations that are not detectable by single-mutation approaches, indicating that standard genotyping significantly underestimates true resistance complexity. The EuResist Network, a European clinical decision-support consortium with retrospective data from 1998 across nine national cohorts, represents the most mature clinically deployed AI resistance prediction system in this dataset.

AI-PRS Platform for TB and Exascale Precision Medicine

The UCLA-developed AI-enabled parabolic response surface (PRS) platform systematically identified near-universal TB drug regimens comprising approved drugs by exploring combinations across a compressed search space — with results described as preclinically superior to the standard regimen in murine models. This platform signals a generalizable methodology: rather than discovering new chemical entities, AI is used to optimise the deployment of existing approved drugs into novel combinations that overcome resistance. According to WHO global TB reports, drug-resistant TB remains one of the most urgent antimicrobial resistance challenges, making AI-optimised regimen design a high-priority translational goal.

At the frontier of the field, a precision medicine initiative from Harvard Medical School — the Proliferation Pathway Network Atlas — proposes patient-specific AI-driven forecasting of resistance targets by mapping parallel proliferation pathway networks, integrating mutation profiles and chromatin accessibility data at an exascale-computing level. This framework represents the most ambitious AI-resistance prediction architecture in the dataset, though it remains at the conceptual and early computational stage.

DDR Inhibition and Efflux Pump Targeting: Preventing and Reversing Resistance

Two distinct mechanistic clusters in the retrieved dataset address resistance through DNA damage response (DDR) pathway inhibition and ABC transporter efflux pump blockade — strategies that aim to prevent resistance emergence rather than simply overcome existing resistance mutations.

AsiDNA and Pan-DDR Pathway Blockade

AsiDNA (Dbait), developed by Onxeo in collaboration with INSERM, is a DNA-mimicking molecule that targets all DNA break repair pathways simultaneously by activating a pan-DNA-repair alarm response. In cyclic treatment protocols, AsiDNA co-treatment abrogated the emergence of resistance to PARP inhibitors niraparib (ovarian cancer) and talazoparib (small cell lung cancer) in cell line and xenograft models. KBM7 haploid cell screens demonstrated a significantly lower frequency of AsiDNA-resistant clones compared to olaparib — a mechanistic finding that distinguishes AsiDNA from conventional PARP inhibitors. The INSERM patent (2020, IL jurisdiction, pending) describes clinical application context, and Onxeo’s 2019 publication presents the experimental basis for clinical translation.

Efflux Pump Inhibition in CML and MDR Tumours

Multidrug resistance mediated by ABC transporters (P-glycoprotein/ABCB1, BCRP/ABCG2) is addressed in retrieved results through P-gp inhibitor combinations and early patent-protected MDR reversal compounds. The University of Coimbra demonstrated that elacridar — a dual P-glycoprotein and BCRP inhibitor — restored imatinib sensitivity in BCR-ABL1-driven, imatinib-resistant CML cell lines (K562-RC, K562-RD) by modulating efflux transporter activity. Both resistant models showed increased BCRP and P-gp activity, confirming transporter-mediated efflux as a primary imatinib resistance mechanism in these lines.

Historically, Merrell Dow Pharmaceuticals filed among the earliest IP addressing the MDR mechanism — pyridyloxazole-2-ones for P-glycoprotein efflux pump inhibition in MDR tumours (1991, AU) and tetraarylethylene compounds for MDR abolition (1997, DE). These patents are now inactive and expired, leaving the efflux pump inhibitor combination space largely open for new IP development. Research published by Nature has documented the structural biology of ABC transporters relevant to MDR reversal strategies, providing a foundation for rational inhibitor design in this space.

High-throughput screening across 3,282 compounds in anaplastic thyroid cancer (ATC) — a disease with no effective therapy — identified CUDC-101 as a first-in-class dual EGFR/HER2/HDAC inhibitor with superior efficacy, inhibiting MAPK signalling and histone deacetylation simultaneously. This NCI/NIH-led discovery exemplifies how multi-target inhibition via a single molecule can address the parallel pathway activation that underlies treatment-refractory disease.

Synergistic Combination Strategies: Evidence and Emerging Directions

Combination therapy for treatment-refractory disease has moved from empirical to data-driven design, with AI and systematic in vitro screening now providing the evidence base for combination selection. The retrieved dataset documents several distinct combination paradigms with meaningful preclinical validation.

NNRTI + NRTI Synergy Against Multi-RTI-Resistant HIV

Combining new NNRTIs (DAPA-2e, DAAN-14h, DAAN-15h) with AZT produced strong synergism against multi-RTI-resistant HIV-1 strains, including those resistant to both NRTI and NNRTI classes. Hebei Agricultural University data suggest this approach may serve as salvage therapy for treatment-experienced patients who have exhausted standard cART options. The DAAN series was also shown to have low nanomolar activity against E138K and E138K+M184I NNRTI-resistant strains, developed at the University of North Carolina.

Dual MET + ERBB Inhibition for Osimertinib Resistance

Dual MET and ERBB blockade has been shown to overcome intratumour plasticity in osimertinib-resistant NSCLC. Amplicon sequencing and digital PCR were used to track resistance mechanisms across treatment timepoints in patient-derived xenografts — a methodology that enables real-time monitoring of resistance evolution and informs sequential combination design. This approach is directly relevant to the clinical challenge of post-osimertinib disease progression, which currently lacks approved treatment options.

Driver Gene-Based Multi-Targeted Combinations

A Budapest/UC Davis approach to customising drug combinations based on each tumour’s driver gene mutation pattern found 25–85% of tested combinations highly synergistic across six cancer cell lines plus primary patient cultures. This range reflects the heterogeneity of driver gene landscapes across tumour types and underscores that driver gene-based combination design significantly outperforms monotherapy. According to data published by NIH, driver gene profiling is increasingly integrated into clinical trial design for refractory cancers, supporting the translational relevance of this approach.

Systematic TB Combination Landscape Mapping

A Boston-based systematic measurement of all two- and three-drug combinations among ten antibiotics across eight lesion microenvironments generated more than 500,000 measurements — feeding ML classifiers that predict in vivo mouse treatment outcomes. This data-generation framework is described as applicable to multidrug-resistant TB and potentially extensible to other drug-resistant infections. The scale of this effort — over half a million data points — signals a new paradigm in which AI is not just predicting combinations but is trained on exhaustive empirical datasets that no human research programme could generate without computational infrastructure. Standards bodies including WHO and NIH have both identified systematic combination testing as a priority for addressing antimicrobial resistance.

Pipeline Signals: What Is Closest to the Clinic

The retrieved dataset is predominantly academic and preclinical, but several clinical and translational signals are meaningful for pipeline assessment. Innovation activity is dominated by academic consortia and publicly funded institutions — NIH, INSERM, Institut Pasteur, CNRS, Tel-Aviv University, UC Berkeley, UCLA, Harvard Medical School, and the Institute for Basic Science (IBS) in South Korea — while commercial entities appear primarily in older MDR reversal IP (now inactive) and isolated clinical-stage compound disclosures.

Abivertinib (AC0010) — Clearest Clinical Signal

Abivertinib, a pyrrolopyrimidine-based irreversible third-generation EGFR inhibitor developed by Hangzhou ACEA Pharmaceutical Research, is described as clinically investigated in China, the United States, and Europe for EGFR T790M-resistant NSCLC. This represents the clearest clinical-stage signal for an SAR-guided resistance-targeting compound in this dataset.

Darunavir — Clinically Validated MDR HIV Backbone

Darunavir/ritonavir combination is positioned as a backbone for multidrug-resistant HIV management, with extensive clinical trial data referenced — pivotal trials in treatment-experienced HIV patients demonstrating virological suppression superior to control PI arms, sustained CD4 count increases, and defined resistance-associated mutation profiles from 2006–2017 US clinical surveillance data. Eleven darunavir resistance-associated mutations and 23 primary PI resistance-associated mutations are characterised in clinical datasets within the retrieved literature.

EuResist Network — Mature Clinical AI Deployment

The EuResist Network clinical decision-support system, released online in 2008, predicts optimal antiretroviral therapy regimens based on patient virological and clinical data — with retrospective enrolment from 1998 across nine European national cohorts. This represents the most mature clinically deployed AI resistance prediction system in this dataset, and a proof of concept that AI-driven resistance prediction can be operationalised at scale in clinical practice. Research infrastructure supporting this system has been documented in peer-reviewed literature and aligns with EMA frameworks for AI-assisted clinical decision support.

AsiDNA and AI-PRS TB — Translational Stage

AsiDNA (Onxeo/INSERM) has a pending patent in IL jurisdiction (2020) describing clinical application context, with experimental validation in cell line and xenograft models. The AI-PRS TB regimens are validated in murine models with results described as much more effective than the standard regimen; human clinical translation is proposed but not confirmed within retrieved results. No retrieved results contain regulatory approval decisions or trial outcome data for AI-discovered targets directly — the majority of AI and computational modalities remain at preclinical or early translational stages.

The overall pipeline signal from this dataset is that academic consortia are generating the foundational target discovery and combination evidence, while commercial translation — particularly for AI-discovered SR rescuer genes and hybrid molecule scaffolds — remains nascent. For drug developers and IP strategists, this represents a window in which foundational academic insights can be translated into proprietary assets before the commercial field becomes crowded. PatSnap’s drug discovery intelligence platform provides the tools to systematically map this landscape and identify white-space opportunities before competitors do.