KIF18A in Aging

KIF18A is a kinesin motor protein that travels along the spindle fibers a dividing cell assembles to draw its chromosomes apart, and that regulates the rate at which the growing ends of those fibers extend and shorten. Loss of the protein is nonetheless compatible with life in the mouse, where animals lacking it are viable and the one confirmed consequence is male infertility, whereas the human evidence runs the other way, in that the constraint band recorded for the gene marks it intolerant of loss of function.

That murine tolerance is what the case for an inhibitor rests on, and it is also the first place the evidence thins, because the genetic case is graded on a constraint band that was never checked against the underlying population database, while the safety read rests on the very mouse work that the aging argument reads in the opposite direction.

Every figure in the card above is reproduced as recorded rather than recalculated here, with identity credited to UniProt and NCBI Gene, structures to PDBe and RCSB PDB, and variants to GWAS Catalog and ClinVar. Species identity is credited to Ensembl Compara and baseline expression to the Human Protein Atlas, each accessed 2026-08-13, whereas the constraint band is the one figure for which no database attribution is recorded.

The paralog table and the stated selectivity concern

The case for selectivity rests on the five kinesin paralogs plotted above, of which the closest, KIF18B, shares 36.4 percent of its sequence with KIF18A, while the remaining four diverge further still, a separation read as sufficient for a molecule that engages one member of the family without engaging the others. A different mitotic kinesin is nonetheless identified as the selectivity concern that governs the program, and that kinesin has no row in the alignment table from which the figure is drawn.

The measured identities and the stated warning therefore concern different proteins, in that the distances above reduce selectivity within the graded family to a medicinal-chemistry problem, whereas the warning places the liability it names outside that set, in a second mitotic kinesin that a molecule aimed at this target is required to spare, because inhibiting it as well would carry back the marrow liability described above. The two statements occupy a single paragraph of the record without being reconciled, and both are carried onto this page in that condition, since choosing between them would assert more than the evidence settles.

Seven experimental structures of KIF18A have been deposited, the sharpest of which is judged suitable for structure-based design, and the site at which the inhibitors in this field are aimed is an allosteric one on the motor domain, the alpha-4 helix, rather than the nucleotide pocket every kinesin possesses.

Because the motor domain is conserved across the kinesin family even where the sequence around it diverges, selectivity against the nearest paralogs cannot be read from the KIF18A structure alone and has to be established by comparison with them, which every compound now in the clinic reports having done. Five relatives are named, and the estimate of how far apart they sit is not internally consistent: a summary figure and a tabulated one contradict each other.

KIF18A: seven listed, by X-ray crystallography and cryo-electron microscopy.

Patent filings against the target

Five patent holders stand on record against this target, as of a search dated 2026-08-13 whose stated account is reproduced beneath the table in the words it was given, although the search itemizes no patent number against any of the five holders it names. The field is judged active at source; the figures available here to set against that judgment are the five holders named in the table below and the filings, where any are recorded, listed beside each of them. Freedom to operate, the right to make and sell a compound without infringing a patent held by another party, is settled by the chemical series a program runs on rather than by the protein it aims at.

About 693 patent documents stand against KIF18A, held by the assignees named below, for none of whom is a patent number or a priority date on record; the count and the holders were retrieved from FreePatentsOnline, accessed 2026-08-13. A dash means no chemical series was recorded for that holder, which is the case for five of five.

Two independent searches, one over ClinicalTrials.gov and one over a curated trials dataset, both conducted on 2026-08-13, returned the same picture: a crowded early-stage oncology field, and nothing whatever outside it.

That constitutes the whole of the human evidence, so that every claim connecting KIF18A to aging rests ultimately on work carried out in patients with advanced cancer, and the record reads that transfer two ways, bridged inside the oncology trials on one reading and rejected as unsupported and mechanistically self-contradictory on the other.

Rows are ordered by stage and then by the number of registry records each program carries, and the square marks the program that has reached the furthest stage; where anything stands on record about a program beyond its registry entry, it is printed beneath the setting.

The one trial with a reported efficacy readout

Everything set out below comes from NCT05902988, Volastra’s study of VLS-1488, whose figures were presented in abstract 3012 at the American Society of Clinical Oncology (ASCO) 2025 meeting, and that presentation is the only occasion on record on which any of these compounds has been given to patients and the resulting numbers put into the public record.

The stage, sponsor, status and enrollment field for this trial were read from the registered landscape, whereas the efficacy and safety figures in the findings column were retrieved separately and are attributed to the 2025 conference abstract named above.

Expression ranking against the clinical record

KIF18A has been scored across thirteen diseases, and the top of that ordering is set here against the clinical record. The ordering itself is not in dispute, whereas the evidential weight that ordering carries remains open, given that an expression fold change and a Phase 1 readout are not the same class of evidence, and given that the two aging entries rank lowest in the scored list.

Composite scores, fold changes and p-values are carried through unaltered, whereas the clinical column records what has reached a patient; fold changes are combined meta-analysis values on a log scale, and p-values are printed here in exponent notation rather than in the superscript form they were scored in.

KIF18A has been scored across a broad field of candidate indications, and the leading entries, grouped by therapeutic area, are dominated by cancers carrying high chromosomal instability, which is precisely the population a KIF18A inhibitor is designed to kill.

Two age-related entries appear within that shortlist, only one of which carries a measured fold change of its own, and both arrive with a reliability flag attached warning that they should not be pursued on the score alone.

Differential expression in oncology and age-related indications

The largest fold change in the table is 0.520, in hepatocellular carcinoma, whereas the only age-related disease it contains is Alzheimer’s disease at 0.121, the smallest positive value the table prints, and that value is attributed to dividing glial cells rather than to any KIF18A mechanism.

Each bar is one disease’s log2 fold change, which is the base-two logarithm of the ratio between KIF18A transcript abundance in diseased tissue and its abundance in matched normal tissue, so that a positive value denotes higher abundance in the diseased state. Three of the fourteen values are negative, and each of those three belongs to a hematological malignancy.

These are the normal-tissue levels against which the fold changes are measured, and in normal tissue KIF18A behaves as a proliferation marker rather than as a broadly expressed gene, given that the three tissues carrying the most of it are the three that divide the most. Baseline abundance in normal human tissue is highest in testis, at 12.8 normalized transcripts per million (nTPM), followed by lymphoid tissue at 8.9 and bone marrow at 7.8.

Indication ranking across the two scoring runs

Every indication graded Go is a cancer, whereas the three indications graded No-go include both age-related entries; the table carries the thirteen indications that were sorted into tiers, together with three further indications that the 1,000-indication run ranked and the tier tables never scored, namely papillary renal cell carcinoma, diabetes mellitus and thrombotic disease.

The columns are arranged as the two runs, either side of the disease name, so that tier, composite score and the first fold-change column belong to the 13-indication run, whereas the second fold-change column, the rank and the call belong to the 1,000-indication run. The wording each run gives a row, together with the points to verify before that row is relied upon, is set out in the second table rather than compressed into the first.

A dash marks a value that was never printed rather than a result that missed a threshold, and every figure comes from the run named in the column note, at the precision that run prints it.

The first two columns are reproduced word for word and keep the spelling they were given, whereas the last column is written here and states how the rest of the evidence bears on the row beside it.

Ranking and verdict for the two age-related indications

Two age-related indications clear the filter applied in the wider run, neurodegenerative disease at rank 40 and diabetes mellitus at rank 37, and both are graded No-go on the ground that their scores arise from network co-embedding rather than from direct evidence, which is to say that they record the position KIF18A occupies within a protein interaction graph rather than the function KIF18A performs in an aged or diabetic tissue. The same conclusion is reached independently by querying aging directly, which returns an attention score of 0.000, and again from a third direction by the 13-indication run, which places both entries alone in its bottom tier, below every cancer it scored.

The finding is recorded within a Red reliability band, and the two age-related indications it concerns are ranked 40 and 37 out of the 1,000 indications that run scored; because the reason behind a grade carries further than the grade itself, the two sentences reproduced below are given in their recorded wording, setting out the basis on which those scores were reached.

A dash marks a field that was not recorded for that row, and the thrombotic indication is retained beside the other two so that the age-related verdicts are not read as the only negative calls the run returned.

Dimension scores returned by the direct aging query

Aging was queried directly rather than read off an indication list, and every dimension that would register a genuine disease association returned zero.

KIF18A is not one of the genes that change with age, given that the aging differential-expression meta-analysis returned 2362 genes at a false-discovery threshold of q below 0.05 and KIF18A was not among them.

Approached from the clinical trial record rather than from the omics data, the evidence returns the same answer, and it is recorded there without qualification.

The interactome community score and the pathway score measure the same network co-embedding named above, and they are the only above-baseline omics scores the aging query returned, both of them being attributable to KIF18A’s position in the mitotic network rather than to anything about aging.

Composite scores and the reproductive-aging contraindication

The 13-indication run sorted its two age-related entries, neurodegenerative disease at 3.5 and reproductive aging at 4.5, into a tier of their own against 9.2 for the leading cancer, and rejected the dual-purpose oncology-aging case for a KIF18A inhibitor outright.

The strongest aging link on record argues against inhibiting KIF18A rather than for it, since the reproductive-aging evidence is read as showing that KIF18A protects oocytes from aneuploidy, which makes inhibition the opposite of the intervention that evidence supports.

The one route left open is indirect and is itself graded speculative: aneuploidy drives cellular senescence, senescence is a hallmark of aging, and KIF18A dysfunction promotes aneuploidy in dividing cells. Every step of that chain concerns cells that divide, which is the objection raised against the aging thesis as a whole.

Statuses are reproduced in their recorded wording, and the second description of that rank is printed in the table of clock streams further down this page. There the KIF18A methylation site stands ninth of 227 features in the mammalian life-history clock, and that position is described at source as a top-ten feature whose coefficient direction is consistent with slower aging. The rank behind the two descriptions is identical and is not itself in dispute, and neither of them states the threshold at which a rank inside a field of a few hundred features becomes a leading one, so both are carried here rather than either being preferred.

Untapped indications in the wider run

The list holds one age-related indication among fourteen cancers, and the three scores that place it there are a protein interaction score, a matrix factorization score and a pathway score, whereas that same indication is graded No-go within the same run.

Names, areas and drivers are reproduced in their recorded wording, and near-duplicate rows are retained rather than merged, so that the ordering is read as it was returned rather than as a deduplicated version of it would appear. The names carry the capitalization of the wider run, which is not the capitalization the narrower run uses for the same diseases.

The same scoring engine was run over KIF18A twice, at two different scales, and the two runs reported different fields. The narrower run scored thirteen indications and sorted them into four priority tiers, whereas the wider run scored a thousand indications across fourteen therapeutic areas and reported a position within that field; at no point were the two sets of numbers reconciled with one another.

Where both runs name the same disease, both positions are carried below and neither is preferred, since there is no principled basis on which to prefer one. The chart draws every indication that the narrower run both placed in a tier and gave a position in the wider field, and the positions it prints beside bladder cancer, clear cell renal cell carcinoma and age-related oocyte aneuploidy are recorded by that run’s tier table alone, given that the wider run’s own tables place none of the three. What the chart shows is that the two orderings disagree: the indication ranked highest in the tier table sits twenty-second in the wider field, whereas the entry that same table places last sits fortieth.

A dash marks an indication that the gene heading the column did not rank at all.

A grade carries little information until the evidence beneath it has been named, and the thirteen indications graded in the 1,000-indication run, a set that overlaps the narrower run’s tiered thirteen without coinciding with it, are not supported to a comparable standard. Some rest on xenograft and patient-derived xenograft models together with active early-phase clinical programs, while others rest on cell-line expression and network inference alone, and the distinction matters more than the ordering the composite score produces.

What follows records the strength of the evidence behind each indication, the features that count in its favor and against it, and the databases from which each figure was drawn, together with the date on which each was accessed.

All thirteen grades come with a reason.

Every retrieval records the date it was made.

The mechanistic case is set out below as it has been constructed, together with the supporting literature adduced for it, and those citations stand as they were recorded rather than verified one by one against the papers they name.

The one failure the record does carry is a 2020 study filed under evidence for inhibition, although the compound it used is a kinesin potentiator that improves mitotic fidelity, and that contradiction is set out with the aging evidence rather than marked in this section.

The two lists hold nine mechanistic arguments and five recorded next steps, and they do not name the indications the same way, so that three of the headings appear word for word in the other list. They are set out below as each list recorded them rather than as pairs, because a pairing inferred from similar names risks filing a next step under an argument about a different indication.

Every sentence in the right-hand column is reproduced as it was recorded, citations included, although none of those citations was checked against the papers themselves.

Every one of these has a recorded result that would rule it out. Both columns are reproduced word for word, apart from two passages shortened where an ellipsis marks the words left out.

The table above is two entries shorter than its source promised: the evaluation undertook to record one next step for every Tier 3 candidate graded Go or Conditional, its own verdict table holds seven such candidates, and it wrote five. Prostate carcinoma and medulloblastoma, both graded Conditional at Tier 3, therefore carry no proposed next step and no result that would rule them out.

The aging argument advanced for KIF18A is a chain of four links: aging tissue accumulates aneuploid cells, those cells turn senescent, senescent cells inflame the tissue around them, and a drug that killed the aneuploid population selectively would therefore break the chain at its first link.

The links are cited unevenly: the passage from aneuploidy to senescence and from senescence to inflammatory signaling rests on published work, whereas the accumulation of aneuploid stem cells with age and the dispensability of KIF18A in healthy stem cells are advanced side by side with no supporting evidence for either. No experiment has yet carried the chain from its first link to its last, a gap acknowledged in four separate places by those who have argued for it.

Confidence was assigned on a five-point scale at source, and each grading is reproduced as it stands rather than recomputed from the claim and the limit printed beside it.

The four arguments were numbered in the order in which they were advanced at source, and that order is preserved here rather than being re-ranked by the weight of the evidence behind each.

Two kinds of measurement are offered as direct evidence that KIF18A is involved in human aging: the weight that epigenetic aging clocks give it, and how its expression changes with age across human tissues.

Both are reproduced here exactly as they were derived, including at the points where the underlying numbers disagree with one another: the clock coefficients carry opposite signs on the same gene between two of the models, and a rise with age in some tissues is read alongside a fall in others as though the two observations supported a single conclusion.

Four discrepancies are reproduced as they stand rather than corrected. KIF18A is stated three times to appear in 4 aging clocks, whereas the summary beneath carries the five rows listed here. The mammalian coefficient appears as +0.6686 in one place and rounded to +0.669 in another, which is one number at two precisions rather than two readings of it. The ZhangBLUP row names its feature as 11 CpG sites and tallies 8 negative against 4 positive, while the detail underneath it runs to 14 rows, 10 negative and 4 positive — three incompatible counts of one set. PASTA and REG carry opposite signs on the identical gene identifier ENSG00000121621 from the same 2025 preprint, and both are counted in favor. One further overlap passes without comment: cg14927277 is both the AltumAge feature and row 13 of the ZhangBLUP table.

Each field is reproduced here as it stood on the date the resource was checked, rather than being re-queried for this page, so an entry added since would not appear.

Every lifespan figure in this table belongs to BubR1 or to p16-positive cell clearance, and none of it is a KIF18A measurement. The last row runs in the opposite pharmacological direction to inhibition, and the same study is elsewhere described as small-molecule inhibition. What the mitotic-kinesin class to which KIF18A belongs does supply is clinical precedent rather than longevity precedent. Ispinesib, an inhibitor of the related mitotic kinesin KIF11, was carried into Phase 2 in melanoma, head and neck squamous cell carcinoma, hepatocellular carcinoma, renal cell carcinoma and prostate cancer without an objective response in any of those trials, and the KIF18A inhibitor AMG 650 completed a Phase 1 study in advanced solid tumors. What that class has never supplied is a lifespan or healthspan measurement of any kind, since no healthy-volunteer study stands on record for any of these agents and no efficacy readout in aging has been reported anywhere in the class.

Were the aging case to be pursued, the shape such a program would have to take is itself the clearest available statement of how far the present evidence actually reaches, and it is set out below.

Every proposal listed begins with an experiment that has never been run, which is not a criticism of the proposals so much as a measurement of the gap they exist to close.

Four development positions are available, three of which command broad agreement: that the oncology program should run, that a standalone anti-aging program should not, and that an inhibitor should not be pointed at reproductive aging, where the evidence establishes that the target is protective. Agreement fails on the fourth, which is whether a single compound can be developed and described as both at once.

That split is left standing in the sections below rather than resolved by assertion, because the measurement that would settle it has not been made in any species. Each position is therefore set out with the reasoning that supports it, and where two lines of reasoning lead to opposite conclusions from the same evidence, both are printed and the point of divergence is named.

Low-cost checks that require no new experiments

Four of the eight matters in dispute on this page, of which seven are still open and one is already settled, can be closed by consulting an existing record or by writing a sentence into a protocol, rather than by running an experiment. They are set out in the order in which they should be undertaken, with the least costly to settle standing first.

1. Read the ASCO abstract behind the clinical activity figure

   The figure rests on one abstract and one denominator, so every appearance of it stands or falls together rather than independently.

2. State the rank threshold the aging clock claim is applying

   The rank itself is printed on both sides of the disagreement, whereas the cut that would make it leading is printed on neither.

3. Say which object the program counts are counting

   A table of named assets that can be checked row by row already exists, and it settles the matter for that class of object.

4. Add senescence, SASP and aneuploidy readouts to the oncology protocols now, by amendment to the studies that are already enrolling

   Of the steps available, it is the only one that converts a trial already under way into evidence bearing on aging rather than on oncology alone.

Nothing argued above is stronger than the material it is drawn from, and that material has three distinct weaknesses which a reader is entitled to meet gathered in one place rather than scattered as qualifications. The case for a role in aging has never been tested directly; several of the figures attached to individual indications do not survive comparison with the tables beneath them; and the conclusion turns on evidence that does not yet exist. Each is set out here in turn, and the part closes with the measurements that would decide the matter in either direction.

The aging case is assembled from association rather than from experiment

The preceding parts assemble an argument from association rather than from experiment, and the distinction is worth stating in full before the individual weaknesses are set out. The accumulation of chromosomally abnormal cells in older tissue, the inflammatory phenotype those cells acquire once they turn senescent, and the selective dependence of such cells on this motor for survival have each been measured. Those measurements were made separately, however, in different systems and for different purposes, and what joins them here is inference rather than any single study that spans them.

Nothing has yet been dosed with an inhibitor of this motor and then followed to an endpoint that measures aging itself, which leaves the step from a coherent mechanism to a demonstrated effect untaken. The curated resources record the same silence, since the target carries a hallmark count of zero and no flag at all in the geroprotector, longevity-gene and trial-registry fields against which it was checked. Whether that silence records an absence of evidence or an absence of curation is itself unsettled, and the two readings are given equal standing here because the fields alone do not decide between them.

The ordering of indications on which much of this page depends inherits the same limit at one remove. A high position in that ordering records how prominently a target-disease pair figures across the scoring dimensions, which is a statement about the volume and recency of the literature behind a pair, the omics co-association measured for it and the funding attached to it. None of those quantities speaks to causation or to clinical validity, and the evidence underneath the ordering is thinner than a position in a ranked list makes it appear. The genetic association has not been fine-mapped, the expression values are taken pre-computed rather than recomputed, and the age-related entries in particular are recorded as having no direct mechanistic support of their own.

Figures printed against indications the underlying tables do not carry

Four of the nine entries registered below arise in the same way, and the mechanism behind them is better stated once than repeated underneath each of them in turn. In every case a differential-expression figure is printed in the scorecard against an indication for which the differential-expression table holds no row of any kind, so that a reader checking the cell against the table it summarizes finds nothing there to check it against.

Two things then happen, and which of them has happened decides how much weight the cell can bear. Sometimes the figure can be traced to a neighboring disease whose row does exist and whose value it reproduces to within the rounding, in which case the number is real and the label attached to it is not. Sometimes no table in either run prints the value against any disease at all, in which case the cell has no ancestor that can be found and is marked here as unsupported.

Nothing has been quietly repaired and no cell has been corrected in passing, which is a deliberate choice rather than an oversight. Each figure appears against the indication it was attached to, at the precision in which it was recorded, and the contents of the underlying table are set beside it. Both the printed value and the thing that stands behind it therefore remain visible at once, and the reader alone is left to decide what to make of the distance between them.

Notes on how the remaining material has to be read

The other five entries are of a different order, in that several of them record disagreements between the underlying tables, yet none impeaches a figure this page itself relies on and none disturbs a conclusion drawn anywhere on it. Each of them nonetheless changes how a row beneath it should be understood by any reader who intends to carry that row somewhere else.

They concern a disease entered twice within one list under two names, two runs that place different kidney subtypes at the same position, a rank left blank in one table and filled in the other, a disease name spelled differently by the accounts that print it, and two readings of one measurement given on a single page and never reconciled. Each is reproduced as recorded rather than tidied, because a corrected spelling, a filled cell or a merged pair of rows would make the underlying material look more orderly than it is, and the orderliness so produced would belong to the transcription rather than to the record it is drawn from.

KIF18A is a kinesin-8 family motor protein whose catalytic motor domain spans residues 11-355, and its tractability as a small-molecule target is settled rather than argued. Seven experimental structures stand on record, the sharpest of them 3LRE at 2.2 ångström, which covers the motor domain and is recorded as suitable for structure-based drug design. A second, 9YMG at 2.41 ångström, captures the motor bound to tubulin and to a non-hydrolyzable ATP analog, and is recorded as providing mechanistic insight into the catalytic cycle. The human genetic evidence is thinner, and it is graded moderate and indirect for reasons the record states plainly. The number of genome-wide association signals at the locus is 0, ClinVar holds 152 variants of which 20 are classified pathogenic, and the constraint band entered for the gene, LOEUF < 0.6, was never checked against the population database it was drawn from. The patent position is recorded as active and is held across five named assignees against none of whom a patent number is registered, so the space is competitive without yet being crowded by granted composition-of-matter claims.

The therapeutic hypothesis this evidence actually carries is an oncology hypothesis, and it rests on chromosomal-instability-selective synthetic lethality, cells carrying an unstable karyotype requiring the motor to complete division while cells that do not are largely indifferent to its loss. That selectivity is quantified as more than 40 percent toxicity in chromosomally unstable cells against less than 20 percent in chromosomally stable and normal cells, which is the whole of the therapeutic window the class is built on. Ten registered programs are counted while every registry entry stands for a program of its own, and the lower of the two tallies reaches at least nine once the single compound holding two of those entries is counted as one asset; 100 percent of them recruit patients whose disease is already advanced, and chromosomal instability is the axis the class selects on in principle, although these studies enroll by tumor histology rather than by any measurement of instability made in the patient. The furthest any has gone is VLS-1488 in NCT05902988, where 7 of 17 evaluable patients showed tumor reduction, reported as 41 percent, alongside 3 partial responses and 6 patients with stable disease, and with no dose-limiting toxicity at any level in 52 patients treated at doses up to 800 mg. Those figures reach this page through abstract 3012 of the 2025 meeting of the American Society of Clinical Oncology and were never traced back to a primary source, and they do not reconcile internally, since 3 partial responses together with 6 patients at stable disease exceeds the 7 credited with tumor reduction out of 17. The only completed study in the class, NCT04293094 in 66 patients, returned its design and its primary outcome measures and no result of any kind, which is why the claim that inhibitors of this motor are tolerated well enough to advance is graded high while the claim of early single-agent anti-tumor activity is graded no higher than low.

The aging hypothesis rests on association rather than on intervention, and the supports it stands on can be named exactly. KIF18A is mapped onto four hallmarks of aging, graded between three and five stars on a five-star scale; it is present in four aging clocks without ranking as a top feature in any of them, which is one of two descriptions this page carries of that rank, the other designating it the leading signal of the mammalian life-history clock at source; and its expression falls with age most steeply in ovary, at r = -0.274 and p = 1.2 × 10⁻⁴ across 193 donors. The remaining support is argument by analogy, taken from mitotic-checkpoint mouse genetics and from senescent-cell clearance, and the only lifespan figure available anywhere in that material, about 25 percent extension of median lifespan, was obtained by clearing p16-positive cells from wild-type aged mice, so that it measures senescent-cell clearance and not KIF18A. What has never been measured is stated with the same precision: the count of published lifespan or healthspan experiments using either an inhibitor of KIF18A or a KIF18A genetic model is 0, the human-aging association count entered against the gene is recorded as 0, and the three curated aging resources checked for it — GenAge, the geroprotector list and the ClinicalTrials.gov aging lists — each return no entry at all. Stated at source, no published evidence links KIF18A directly to aging biology, cellular senescence, or neurodegeneration, and no senescence or senescence-associated secretory phenotype measurement has been made on the motor itself, which is why the proposition that an aneuploidy-selective senolytic works in human beings for aging is graded very low, its closest precedent belonging to a different senolytic class, dasatinib with quercetin.

In the narrower run, thirteen indications were scored on a composite and sorted into four tiers, eleven of them cancers and two age-related, and both age-related entries were placed in the lowest tier: age-related oocyte aneuploidy and reproductive aging scored 4.5 in tier 4 and carries no rank of its own, the number printed beside it belonging to infertility, while neurodegenerative disease scored 3.5 in tier 4 and stands 40th. The strongest oncology entry, high-grade serous ovarian cancer, scored 9.2 in tier 1, and the weakest, clear cell renal cell carcinoma, scored 4.8 in tier 3, so that every one of the eleven cancers scored above both age-related entries and the margin between the two classes is unbroken. The differential expression beneath those positions separates as sharply as the composites do, hepatocellular carcinoma carrying a log2 fold change of 0.520 at p = 7.08 × 10⁻⁷¹ and gastric carcinoma 0.493 at p = 4.60 × 10⁻⁸⁶, whereas the one age-related measurement, entered as Alzheimer’s disease, carries 0.121 at p = 1.00 × 10⁻² across 21 datasets, the smallest positive fold change among the fourteen diseases measured. In the wider run, across 1,000 indications in 14 therapeutic areas scored on 23 combined metrics, the two age-related positions whose origin is stated outright, diabetes mellitus at 37 and neurodegenerative disease at 40, are attributed at source to network co-embedding rather than to any measurement of KIF18A in those diseases, and both are graded no-go.

One age-related setting has been measured directly, and the measurement runs against the drug rather than behind it, in that motor-domain variants of KIF18A, T273A among them, prematurely raise oocyte aneuploidy in humans and in mice, the recorded human case being a 25-year-old homozygous carrier whose oocytes were 45 percent aneuploid. Reduced KIF18A function therefore accelerates aging in the one tissue where a direct measurement exists, which is why reproductive aging is graded no-go and pharmacological inhibition is recorded as contraindicated there, notwithstanding that the same finding is elsewhere carried as support for the aging case and graded four of five stars in that role. The strongest direct link on record between this target and an aging tissue is thus an argument against inhibiting it, which is a different position from an absence of evidence and a considerably harder one to move.

What follows for the sequencing of any program is a consequence of where the evidence sits rather than a preference between two ambitions. The oncology indication carries seven structures, a patent field recorded as active, ten registered programs, a quantified selectivity window and a first human efficacy readout, whereas the aging indication carries a composite of 3.5, a log2 fold change of 0.121 at p = 1.00 × 10⁻² and a count of interventional experiments that stands at 0, so that the first supports a program and the second supports only a measurement. The whole of the comparison reduces to one sentence: KIF18A is a well-precedented oncology target with a quantified selectivity window and an early single-agent signal in chromosomally unstable tumors, and an unevidenced aging target whose two age-related indications scored beneath every cancer they were ranked against.

The measurement that would overturn that sentence is equally specific, since the step recorded as the cheapest available writes senescence, SASP and aneuploidy readouts into the oncology protocols by amendment, and it is the only step that converts a trial already recruiting into evidence about aging. Two outcomes follow from that single addition, and they divide the question cleanly: a fall in senescent-cell burden or in senescence-associated secretory phenotype markers in treated patients would move the aging case from association to intervention for the first time, whereas no movement in either would leave the aging indication exactly where the composite scores already place it, beneath every oncology indication scored. A lifespan or healthspan experiment in any organism would settle the matter outright, and the number of such experiments now standing on record is 0.

Target characterization drew on UniProt and NCBI Gene for sequence and domain architecture, on PDBe and the RCSB Protein Data Bank for experimentally determined structures, and on Ensembl Compara for paralog relationships and pairwise sequence identity, while human genetic evidence was taken from the GWAS Catalog and ClinVar and chemical and intellectual-property tractability from ChEMBL and FreePatentsOnline.

Disease association and indication ranking were produced with PandaOmics, which scored the target across the disease-indication space on 23 metrics in three families, those being curated disease omics datasets, biomedical literature volume with its recency and citation impact, and grant funding, each normalized between zero and one and combined into one composite rank across the full indication field. The four priority tiers printed on this page were not returned by that ranking but assigned by the narrower run, which banded a composite score of its own in which chromosomal-instability prevalence carried the heaviest weight and the ranking position entered as one input among several. Differential expression was derived from the PandaOmics expression meta-analysis, tissue-level and age-related expression from the Human Protein Atlas and GTEx, pathway and interaction membership from STRING v12.0 and Reactome, registered clinical activity from ClinicalTrials.gov through version 2 of its application programming interface, and supporting primary literature from PubMed.

Primary literature cited

Six published studies carry the core of the argument set out above, and they divide into three groups according to the role each one plays in it. The first establishes the synthetic-lethal vulnerability on which the oncology program is built, the middle four supply the chain that runs from chromosome missegregation through senescent-cell burden to lifespan, and the last is the only one of the six to test a small molecule against any link in that chain. Further primary work is cited by author and year at the point in the text where each result bears on the argument, so this list records that core rather than the complete literature cited here.

1. Cohen-Sharir, Y. et al. (2021) Aneuploidy renders cancer cells vulnerable to mitotic checkpoint inhibition. NaturePMID 33505028
2. Baker, D. J. et al. (2004) BubR1 insufficiency causes early onset of aging-associated phenotypes and infertility in mice. Nature GeneticsPMID 15208629
3. Baker, D. J. et al. (2011) Clearance of p16Ink4a-positive senescent cells delays ageing-associated disorders. NaturePMID 22048312
4. Baker, D. J. et al. (2013) Increased expression of BubR1 protects against aneuploidy and cancer and extends healthy lifespan. Nature Cell BiologyPMID 23242215
5. Baker, D. J. et al. (2016) Naturally occurring p16(Ink4a)-positive cells shorten healthy lifespan. NaturePMID 26840489
6. Barroso-Vilares, M. et al. (2020) Small-molecule inhibition of aging-associated chromosomal instability delays cellular senescence. EMBO ReportsPMID 32134180

Glossary of terms

Aneuploidy

Having the wrong number of chromosomes in a cell. A healthy human cell carries forty-six; a cell that has gained or lost one is aneuploid. It accumulates with age and is close to universal in the tumors these programs target.

Chromosomal instability

A cell’s tendency to keep gaining and losing chromosomes each time it divides. It is the process; aneuploidy is the result. The trials described here choose patients by tumor type instead of by measuring it. Also written as: CIN, CIN-high, CIN-low

Kinesin

A family of motor proteins that walk along the internal fibers of a cell, carrying cargo or adjusting the fibers themselves. KIF18A is one of them, and the part that does the walking is called the motor domain.

Microtubule

The stiff hollow fibers a cell builds to pull its chromosomes apart when it divides. KIF18A walks along these and controls how fast their ends grow and shrink.

Synthetic lethality

When losing either of two things is survivable but losing both is not. Here it is the argument for the whole target: normal cells tolerate the loss of KIF18A, and cells whose chromosomes are already unstable do not.

Senescence

A state in which a damaged cell stops dividing but does not die, and instead persists in the tissue, releasing inflammatory signals. A senolytic is a drug that kills such cells. Also written as: senescent cells, senolytic

Senescence-associated secretory phenotype

The mixture of inflammatory proteins a senescent cell releases, which is how a single such cell degrades the tissue around it. Also written as: SASP

Epigenetic clock

A statistical model that estimates a person’s age from chemical marks on their DNA. The weight such a model assigns to KIF18A is frequently advanced as evidence that the gene participates in aging, an inference whose limits are examined below.

Log2 fold change

How much more, or less, a gene is switched on in diseased tissue than in healthy tissue, on a doubling scale. A value of one means twice as much; a negative value means less. Also written as: logFC

Geroscience endpoint

A measurement in a clinical trial that is about aging itself rather than about one disease, such as a biological-age estimate or a count of senescent cells. No KIF18A trial carries one.

Patient-derived xenograft

A piece of a patient’s tumor grown in a mouse, used to test whether a drug works against that particular tumor before trying it in people. Also written as: PDX

Companion diagnostic

A test used alongside a drug to decide who should receive it. For this target it would be a measurement of how unstable a tumor’s chromosomes are, and no threshold for it has been fixed.