Cracking the SIK2 Code: AI Unlocks a New Era in Oral Immunology

BOSTON – November 10, 2026 – For the past decade, the multi-billion-dollar immunology market has been defined by a frustrating paradox. Biologic therapies, while highly effective for conditions like rheumatoid arthritis and ulcerative colitis, require cumbersome injections or infusions and often lose their efficacy over time. Conversely, the push for convenient oral alternatives has been marred by safety compromises, most notably the black-box warnings attached to pan-JAK inhibitors due to severe cardiovascular and oncological risks. The industry has been searching for a holy grail: an oral small molecule that offers the targeted safety of a biologic with the convenience of a daily pill.

Today, that search enters a critical new phase. Nimbus Therapeutics, a Boston-based computational drug discovery pioneer, announced the first-in-human dosing of NDI-225673, an investigational oral therapy for immune-mediated inflammatory diseases. Evaluated in a Phase 1 clinical trial among healthy volunteers, the compound aims to establish safety, tolerability, and pharmacokinetics. But the true industry intrigue lies in its target: salt-inducible kinase 2 (SIK2).

By deploying an advanced suite of artificial intelligence, machine learning, and physics-based structural modeling, the biotech firm has engineered a molecule that claims to thread one of the most notoriously difficult needles in molecular biology. If successful, it could fundamentally rewrite the standard of care for millions of patients, shifting the therapeutic focus from merely suppressing inflammation to actively repairing damaged tissue.

Beyond Cytokine Suppression: The Dual-Action Mandate

To understand the commercial and clinical potential of NDI-225673, one must look at the biological limitations of current blockbuster drugs. Most approved therapies for autoimmune diseases operate as blunt instruments, targeting single inflammatory cytokines like TNFα or IL-23. While this suppresses the immediate immune flare-up, it rarely addresses the underlying tissue damage, leading to a cycle of relapse and remission.

SIK2 represents a paradigm shift. As a key regulator of transcriptional programs in myeloid cells such as macrophages, SIK2 does not just control a single inflammatory pathway. In preclinical studies, inhibiting SIK2 suppressed a broad orchestra of key inflammatory drivers simultaneously—including TNFα, IL-6, IL-23, and IL-1β.

More importantly, selective SIK2 inhibition triggers a secondary, restorative biological response. Preclinical models demonstrated that targeting this specific kinase increases the production of IL-10, a potent anti-inflammatory cytokine, alongside critical tissue-repair mediators like amphiregulin (AREG) and thrombospondin-1 (THBS1). In mouse models of colitis, this dual action not only reduced disease activity but physically improved gut barrier integrity markers such as mucin and occludin.

“Many patients with immune-mediated inflammatory diseases, such as ulcerative colitis and rheumatoid arthritis, do not reach durable remission on current therapies,” said Peter Tummino, Ph.D., President of Research and Development at Nimbus. “Our goal with selective SIK2 inhibition is to go beyond suppressing individual inflammatory cytokines by engaging biological pathways that regulate inflammation and tissue repair. NDI-225673 represents years of research to achieve a high degree of SIK isoform selectivity, and this Phase 1 study gives us the opportunity to begin testing the SIK2 hypothesis directly in humans.”

Cracking the Isoform Conundrum

The scientific rationale for SIK inhibition has been known for years, but pharmaceutical companies have repeatedly stumbled over the "isoform conundrum." The SIK family comprises three closely related kinases: SIK1, SIK2, and SIK3. Because their molecular structures are nearly identical, early attempts to drug the pathway resulted in pan-SIK inhibitors that indiscriminately blocked all three.

The clinical consequences of this lack of precision are severe. SIK1 inhibition is directly associated with cardiovascular toxicity, a non-starter in chronic disease management. Meanwhile, SIK3 loss of function has been linked to severe skeletal abnormalities and fertility-related defects. To safely harness the therapeutic power of SIK2, a drug must bind to it while completely ignoring its molecular siblings.

This is where the traditional trial-and-error approach of medicinal chemistry fails, and where modern computational biology proves its worth. To achieve the necessary precision, researchers utilized proprietary high-resolution protein-ligand co-crystal structures. By integrating physics-based computational methods, quantum mechanics, and iterative rounds of machine-learning predictions, the design team was able to identify minute, unique binding pockets specific only to SIK2.

Industry insiders note that achieving this level of selectivity—reportedly greater than 300-fold over SIK1 and SIK3 in preclinical assays—requires an immense orchestration of digital tools. By leveraging virtual reality environments to visualize molecular dynamics and utilizing bioinformatics platforms to analyze complex multi-omics data, computational chemists can simulate millions of interactions in silico before synthesizing a single compound in the lab. This approach drastically reduces the time and capital required to de-risk highly complex targets.

The Computational Engine Validated

For business leaders monitoring the biopharma landscape, the advancement of NDI-225673 is more than a clinical milestone; it is a stress test of the AI-native drug discovery business model. Skeptics of AI in pharma often point to a lack of late-stage clinical validation. However, the company behind this novel SIK2 inhibitor has already provided one of the industry's most compelling proof points.

In late 2022, the firm reported highly positive Phase 2b data for a highly selective allosteric TYK2 inhibitor, another notoriously difficult immunology target. Within months, Takeda acquired the program—now known as TAK-279—for a staggering $4 billion upfront and up to $2 billion in commercial milestones. That transaction validated the computational platform's ability to generate best-in-class, highly selective oral molecules that Big Pharma desperately needs to replenish aging pipelines.

Rather than resting on its laurels, the firm has reinvested that capital to advance a wholly-owned internal pipeline. Alongside the SIK2 program, the company is progressing NDI-219216, a Werner syndrome helicase (WRN) inhibitor for solid tumors, and a novel HPK1 inhibitor. This aggressive advancement strategy illustrates a broader industrial shift: specialized computational biotechs are no longer just discovery engines for larger pharmaceutical companies; they are evolving into fully integrated clinical-stage powerhouses capable of taking highly complex assets deep into human trials.

The Future of the Immunology Market

As the Phase 1 trial of NDI-225673 progresses, the broader pharmaceutical industry will be watching closely. The competitive landscape for oral immunology is intensifying rapidly. Major players are advancing oral IL-23 receptor antagonists, next-generation TYK2 inhibitors, and novel mechanisms like RIPK1 inhibitors.

Competitors have previously explored the SIK pathway with mixed results. Galapagos, for instance, advanced a SIK2/3 inhibitor into early clinical trials before halting its progress, pivoting instead to other selective SIK candidates in its pipeline. The historical difficulty of drugging this family of kinases means that any positive safety and pharmacokinetic data from the current Phase 1 trial will be viewed as a major technological triumph.

If the computational design holds up in human subjects—proving that SIK2 can be isolated without triggering the cardiovascular or skeletal liabilities of SIK1 and SIK3—it will unlock an entirely new therapeutic class. For healthcare systems burdened by the high costs and infrastructure demands of biologic infusions, a safe, oral pill that actively promotes mucosal healing and joint repair would be transformative.

The next industrial revolution in pharmaceuticals is not just about discovering new biological targets; it is about having the computational precision to hit them safely. As algorithms and artificial intelligence become inextricably linked with molecular engineering, the barriers to drugging the "undruggable" are falling, promising a new era of systemic, highly targeted interventions for the world's most complex chronic diseases.