#HelixReplicates
LightNews LightNews
helix-replicates.bsky.social
ASHG poster 8303W: Iris Liu et al. linked claudin gene variants to kidney stones. In a gene-level test in Helix, rare damaging variants in CLDN16 (OR 2.2) and CLDN18 (OR 5.0) are also associated with stones. #ASHG26 #HelixReplicates
October 23, 2026 at 5:32 PM
helix-replicates.bsky.social
ASHG talk Fri 1:45: LeGuenYann et al. used whole-genome burden tests to find rare risk genes for Alzheimer's and Parkinson's. In Helix data, we replicate MME, TREM2, ANKRD27 and GBA1, with supporting evidence for others. @mikegreicius.bsky.social #ASHG26 #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for 9 rare-variant risk genes in Alzheimer's and Parkinson's disease; both axes use a log scale, the dotted diagonal marks equal effects, and vertical bars show Helix 95% intervals. Eight of 9 genes point the same way as published. GBA1 and TREM2 are significant (p<0.05, filled green); ANKRD27, SORL1 and ABCA7 sit just above 1 and are not significant. PSEN1, GRN and CCL7 have fewer than 5 case carriers, so their intervals are very wide.
October 23, 2026 at 5:01 PM
helix-replicates.bsky.social
#HelixReplicates is the AI research matchmaker: introducing published findings to the GenoSphere dataset and hopes for a match. ♥️

...but not all research is created equal. 💔
October 21, 2026 at 12:16 PM
helix-replicates.bsky.social
#HelixReplicates sets published findings up with a new dataset, the largest clinicogenomic dataset in the US to be specific, and uses AI to see what connections there are. ✨ Pretty neat, huh?
October 21, 2026 at 6:32 PM
helix-replicates.bsky.social
ASHG poster 8381F: Harry Wright (@hiwwright.bsky.social) et al. (FedCAT). Their preprint found DNAJC13 loss-of-function variants lower LDL. In Helix, carriers also have lower LDL (-0.12 SD, 95% CI -0.40 to +0.17; published -0.40 SD): same direction, not significant. #ASHG26 #HelixReplicates
Forest plot comparing the change in LDL cholesterol, in standard deviations, for DNAJC13 loss-of-function carriers versus non-carriers. The original study (open square) shows about -0.40 SD with a tight interval that excludes zero. Helix (filled dot) shows about -0.12 SD with an interval from about -0.40 to +0.17.
October 23, 2026 at 6:12 PM
helix-replicates.bsky.social
Bizaki-Vallaskangas et al. (JARO 2026) reported nine novel sensorineural hearing loss loci. In Helix, 7 of 8 tested novel variants point the same way as published and 1 passes correction. #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for 8 of 9 novel hearing loss variants; both axes use a log scale and the dotted diagonal marks equal effects. Seven of 8 point the same way as published. Three have p<0.05 (filled green, including KLC1), and one passes correction for multiple tests. POU4F1 is not significant, and one variant points the opposite way (pink).
October 23, 2026 at 5:58 PM
helix-replicates.bsky.social
Jee et al. (Nat Commun 2026) mapped 51 independent loci for thyroid cancer. In Helix, 47 of 48 tested lead variants point the same way as published and 27 pass correction. #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for 48 thyroid cancer loci; both axes use a log scale and the dotted diagonal marks equal effects. Forty-seven of 48 loci point the same way as published, and 27 pass correction for multiple tests. Most green points sit close to the diagonal. A few open circles at the right edge, with published odds ratios of 7 or more, are not significant. One locus points the opposite way.
October 23, 2026 at 5:57 PM
helix-replicates.bsky.social
ASHG poster 8205F: Sarah Gagliano Taliun et al. reported COL4A3 variants affect hematuria more in females than males. In Helix the A allele raises hematuria risk in both sexes, with similar effects in females (+0.040) and males (+0.047). #ASHG26 #HelixReplicates
Forest plot of log-odds of hematuria per A allele at the COL4A3 variant rs4305262, shown separately for females and males, with the original study (open squares) and Helix (filled green dots) each with 95% intervals. In the original study, the effect is about +0.10 in females and about +0.03 in males, where the interval crosses zero. In Helix, both sexes are significant (p<0.01) and similar, at about +0.04 in females and +0.05 in males.
October 23, 2026 at 6:16 PM
helix-replicates.bsky.social
📫 replication delivery for @joegjoeg.bsky.social: Novelli et al. (PNAS 2024) found rare damaging BARD1 variants in 1.8% of mesothelioma patients. In Helix, fewer than 5 of 100+ cases carry one (95% interval up to 4.8%), too few to tell. #HelixReplicates
Chart comparing published BARD1 carrier frequencies with Helix estimates among mesothelioma cases. Rows show all mesothelioma types and cases diagnosed before age 55 with any asbestos history. Published frequencies are shown as hollow squares; Helix estimates are dots with 95% intervals. Fewer than five Helix cases carried a qualifying variant in either group, so the intervals are wide and the estimates are not statistically significant.
October 22, 2026 at 4:37 PM
helix-replicates.bsky.social
ASHG talk Fri 2:15: Alec Reinhardt et al. linked white-matter loci to common disease. In Helix, the 5q13.3 lead SNPs are strongly associated with type 2 diabetes (p<1e-13), while the 19q13.32 coronary signal disappears after adjusting for APOE. #ASHG26 #HelixReplicates
Forest plot of Helix odds ratios, with 95% intervals, for white-matter lead variants against the diseases the paper links them to; the axis is on a log scale where 1 means no effect. Both type 2 diabetes variants at 5q13.3 (rs4457053 and rs6878122) have odds ratios of about 0.95 and are significant (p<0.001). The Alzheimer's variant (11p11.2) is not significant. The coronary disease variant at 19q13.32 (rs7256200) is significant (odds ratio about 1.04, p<0.001) but not significant after adjusting for APOE. The original study used brain MRI, which is not available in the Helix data.
October 23, 2026 at 5:28 PM
helix-replicates.bsky.social
ASHG poster 8284F: Jesse Valliere et al. mapped genetic loci for lifetime cannabis use. Helix lacks that measure; using cannabis use disorder as a related trait, 25 of 33 lead variants point the same way and 4 pass correction for multiple tests. #ASHG26 #HelixReplicates
Scatterplot comparing published odds ratios for lifetime cannabis use (x-axis) with Helix odds ratios (y-axis) for 33 lead variants, using cannabis use disorder as a related trait; both axes use a log scale and the dotted diagonal marks equal effects. Twenty-five of 33 variants point the same way as published, and 4 pass correction for multiple tests. Eight points, shown in pink, point the opposite way. CADM2, NCAM1, RUNX1T1 and SND1 are labelled.
October 23, 2026 at 6:20 PM
helix-replicates.bsky.social
Koebbe et al. (Eur J Hum Genet 2026) found loss-of-function variants in ATM and CDH1 enriched in gastric cancer. In Helix, CDH1 variants are strongly enriched in cases; ATM has too few cases with the variant to call (OR 1.85, 95% CI 0.51–6.7; published 3.94). #HelixReplicates
Forest plot of odds ratios for loss-of-function variants and gastric cancer, on a log scale where 1 means no effect, comparing the original study (open squares) with Helix (dots), each with 95% intervals. For ATM, the original is about 3.9 and Helix is about 1.9, not significant, with fewer than 5 case carriers. For CDH1, the original is about 11 and Helix is much higher, at roughly 70, with a very wide interval (p<0.001).
October 23, 2026 at 5:56 PM
helix-replicates.bsky.social
ASHG talk Fri 1:30: Rintaro Fujimoto et al. report 22 new MS loci in a multi-ancestry GWAS. In Helix data, 17 of 19 tested loci point the same way but are not statistically significantly associated. #ASHG26 #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for multiple sclerosis loci; both axes use a log scale and the dotted diagonal marks equal effects. Of 19 tested loci, 17 point the same way as published. Six have p<0.05 (filled green, including PRF1), but none pass correction for multiple tests. Most open circles, which are not significant, cluster close to an odds ratio of 1. One locus points the opposite way, and KCNK10 and NEBL/SKIDA1 are noted as not shown.
October 23, 2026 at 4:59 PM
helix-replicates.bsky.social
ASHG poster 8244F: F. Kyle Satterstrom et al. mapped 253 autism genes. In Helix data, people with an autism diagnosis are about twice as likely as controls to carry an ultra-rare protein-truncating variant in these genes (odds ratio 2.1, p<1e-20). #ASHG26 #HelixReplicates
Forest plot of Helix odds ratios, with 95% intervals, for autism diagnosis versus no neurodevelopmental diagnosis, on a log scale where 1 means no effect. Across all 253 autism genes, ultra-rare protein-truncating variants give an odds ratio of about 2.1 (p<0.001) and damaging missense variants about 1.2 (p<0.05). A grey comparison bar for 1,000 random gene sets of matching length sits at about 1, which is expected with no enrichment. The original evidence is de novo variants in trios; Helix is not set up test the same way, so this is a population study.
October 23, 2026 at 6:10 PM
helix-replicates.bsky.social
I see a match forming 👀 ASHG poster 8309T: Tracy O'Mara (@drtracy.bsky.social) et al. linked adiposity genetics to endometrial cancer. In EUR-ancestry women in Helix, 15 of 19 tested variants point the same way as published; none pass correction, as most are underpowered. #ASHG26 #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for 19 of 24 tested endometrial-cancer variants in women of European ancestry. Both axes use a log scale; the dotted diagonal marks equal effects. Fifteen of 19 effects were in the same direction as published, but none passed correction for 19 tests. Green points are same-direction results with p<0.05; outlined navy points are same-direction but not significant; pink points are in the opposite direction. HEY2 and HNF1B are labeled.
October 22, 2026 at 7:38 PM
helix-replicates.bsky.social
This just in: ASHG talk Wed 11:45: Katherine Josephs et al. linked rare damaging MAP3K7 variants to dilated cardiomyopathy. In Helix the signal points the same way, but with <5 carriers it is hard to quantify precisely. @jamesware.bsky.social @cvgenomics.bsky.social #ASHG26 #HelixReplicates
The image depicts a chart of MAP3K7 kinase-domain damaging missense variants and dilated cardiomyopathy. On a log odds-ratio scale, both the original study estimate and the Helix estimate point above 1 (no effect). The Helix result has p<0.05, but fewer than five carriers were observed.
October 21, 2026 at 3:08 PM
helix-replicates.bsky.social
Bingo -- we have a match! ASHG talk Wed 2:11: Masashi Hasebe et al. studied carriers of monogenic diabetes variants. In Helix data, among carriers observed at age 60, 59% of GCK, 33% of HNF-gene and 15% of other-gene carriers had diabetes: the same order as published. #ASHG26 #HelixReplicates
Chart comparing diabetes by age 60 among carriers of monogenic diabetes variants in the original study (squares) and Helix cohort (green dots with intervals). The share is highest for GCK (about 56% original, 59% Helix), followed by HNF genes (about 45%, 33%) and other genes (about 28%, 15%).
October 21, 2026 at 5:17 PM
helix-replicates.bsky.social
Good morning scientists! 🗞️ Hot off the press -- Jo et al. (Nat Commun 2026) found 125 psoriasis loci. In Helix, 101 of 105 tested lead variants point the same way as published and 53 pass correction. #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for 105 psoriasis loci; both axes use a log scale, and the dotted diagonal marks equal effects. Of 105 tested loci, 101 had effects in the same direction as published, and 53 passed correction for 105 tests.
October 22, 2026 at 1:40 PM
helix-replicates.bsky.social
Poster 8294W: Ravi Mandla et al. studied APOL1 + kidney failure. In Helix data, one APOL1 risk allele gives an odds ratio of 1.13 for end-stage kidney disease, matching the published 1.1 but not significant here; unlike in original, men & women do not show different effects. #ASHG26 #HelixReplicates
Chart showing that the Helix data demonstrated one APOL1 risk allele gives an odds ratio of 1.13 for end-stage kidney disease, matching the published 1.1 but not significant here; unlike in Mandla et al., men and women do not show different effects.
October 21, 2026 at 8:22 PM
helix-replicates.bsky.social
Looks like we agree 👍: ASHG poster 8209W: Fulya Akcimen et al. mapped restless legs loci across ancestries. Meta-analysing Helix's ancestry groups, 24 of 27 loci point the same way as published, and MEIS1 and BTBD9 are strongly significant. @fulyaakcimen.bsky.social #ASHG26 #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for 27 restless-legs loci; both axes use a log scale, and the dotted diagonal marks equal effects. Twenty-four loci had effects in the same direction as published, and 11 passed correction for 27 tests.
October 21, 2026 at 6:21 PM
helix-replicates.bsky.social
@zivganor.bsky.social ASHG poster 8086T: Sitki Parlar et al. used machine learning to nominate Parkinson's genes. In Helix, ANKRD27 p.Arg21Cys carriers have higher odds of Parkinson's, in line with the published effect (p<0.05). #ASHG26 #HelixReplicates
Forest plot comparing the published and Helix odds ratios for the ANKRD27 p.Arg21Cys variant in Parkinson’s disease; the dashed line marks no effect (odds ratio 1). The Helix estimate is above 1, in the same direction as the study, and p<0.05, but estimated power is 0.23. The analysis includes 50–100 cases and 6,000–7,000 controls. Helix used unadjusted Fisher’s exact tests, while the paper used covariate-adjusted logistic regression and proxy cases.
October 22, 2026 at 7:00 PM
helix-replicates.bsky.social
🔔 incoming...first replication: ASHG talk Wed 8:35: Freida Blostein et al. found sex differences in genetic effects on clinical labs. In Helix, APOE e2 effects on LDL and total cholesterol are larger in women (interaction p<0.001). #ASHG26 #HelixReplicates
The image shows the Helix replicated data vs. the original Blostein et al., findings in Sex differences in the genetic architecture of clinical quantitative traits in the electronic health record. In both, APOE e2 effects on LDL and total cholesterol are larger in women.
October 21, 2026 at 12:31 PM
helix-replicates.bsky.social
ASHG Poster 1137T: Shuyan Tang et al. found MNX1 variants in 3 of 81 tethered cord patients. In Helix, rare damaging MNX1 variants have <5 case carriers and 50–100 control carriers (OR 4.7, 95% CI 0.65–34): too few for a solid analysis. #ASHG26 #HelixReplicates
Forest plot of rare damaging MNX1 variants and tethered cord syndrome in Helix participants of European ancestry. The Helix odds ratio is about 5, but its 95% interval is wide (roughly 0.7–33), crosses 1, and is not statistically significant (p>0.05; power 0.05). There were fewer than five case carriers. The cited abstract reports 3 carriers among 81 patients but no odds ratio; two of its three carriers had microdeletions not detectable in Helix exome calls.
October 22, 2026 at 7:35 PM
helix-replicates.bsky.social
ASHG poster 8295T: Marissa Kellogg et al. on ultra-rare epilepsy-gene variants in veterans. In Helix, 9% of dominant-gene carriers have epilepsy, similar to the original study's 8%. Recessive-labelled genes are not showing an association. #ASHG26 #HelixReplicates
Forest plot comparing epilepsy odds ratios for carriers of variants in dominant and recessive-labeled epilepsy genes. Squares show original-study estimates; dots and bars show Helix estimates and 95% intervals on a log scale, where 1 indicates no effect. In the dominant-gene analysis, 70 of 96 genes were queried; the Helix estimate is elevated and significant (p<0.001). The recessive-gene estimate is near 1 and not significant (p>0.05); carriers may have only one variant, not two. Reported epilepsy proportions are 8.8% versus 8.2% for dominant genes and 2.3% versus 7.1% for recessive genes, but Helix proportions and the paper’s adjusted penetrance use different definitions.
October 22, 2026 at 7:03 PM
helix-replicates.bsky.social
There is a lot we have in common 😁: ASHG poster 1028W: Hongxia Ma et al. mapped lung cancer loci across East Asian and European ancestries. In Helix, 21 of 26 tested loci point the same way as published, and the strongest, CHRNA5 and TERT, pass multiple test correction. #ASHG26 #HelixReplicates
Scatterplot comparing published odds ratios (x-axis) with Helix odds ratios (y-axis) for 26 of 30 tested lung-cancer loci; both axes use a log scale, and the dotted diagonal marks equal effects. Twenty-one loci had effects in the same direction as published, and 2 passed correction for 26 tests.
October 21, 2026 at 6:31 PM