1/50
Looks like no tags are added yet.
Hypothesis (one line)
Mutation type predicts how much cancer risk, and which cancer
The gap
Past studies disagreed on which mutation type is worse; none compared types in a large directly tested cohort
The problem in the clinic
Every LFS carrier gets the same surveillance
Total people tested
8,159
P/LP carriers
5,034
Tested negative (the floor)
2,925
Truncating carriers (count)
1,041
Other missense carriers (count)
2,515
R337H carriers (count)
1,080
Truncating variant
Protein cut short: nonsense, frameshift, splice-site or large deletion; no full-length protein
Missense variant
One amino acid swapped for another
R337H
Brazilian founder missense, arginine to histidine at codon 337, in the tetramerization domain (outside DNA-binding)
Reference group for every hazard ratio
Other missense
Why other missense is the reference
Largest (2,515) and most typical group, so the most stable comparison
Methods (one line)
Kaplan-Meier risk curves and Cox hazard ratios vs other missense, clustered by family, rerun with probands removed
Why cluster by family
Relatives share the same variant, so they are not independent
Proband
First person tested in a family, usually because they already had cancer
Probands-removed result
Everything held: a check, not a correction
Kaplan-Meier curve shows
Share who have had that cancer by each age, counted from birth
Hazard ratio means
Rate of the cancer at any given age vs other missense; above 1 more, below 1 less
HR whose interval crosses 1 means
No detectable difference
Breast HR, truncating vs other missense
1.72 (1.45 to 2.03)
First-cancer HR, truncating vs other missense
1.61 (1.41 to 1.84)
Brain HR, truncating vs other missense
1.06 (0.79 to 1.42): no difference
Brain HR, R337H vs other missense
0.32 (0.21 to 0.48): about a third
Breast HR, R337H vs other missense
0.65 (0.55 to 0.77)
First-cancer HR, R337H vs other missense
0.59 (0.52 to 0.66)
Breast risk by 60: truncating vs R337H
79% vs 56%
Brain risk by 60: other missense vs R337H
11% vs 4% (11.3% vs 3.8%)
First LFS cancer risk by 60, truncating
78.2%
Finding 1
Truncating highest for breast and first cancer
Finding 2
R337H lowest in all three outcomes
Finding 3
Brain is the exception: truncating and other missense alike
Why is 79% not a real-world risk
Families get tested after cancer shows up, no delayed entry, no competing risk, so absolute curves run high
What is the real finding
The comparison between groups, not the absolute percentages
Why brain, breast, first cancer
Brain: summer lead (LoF/DN enriched), Amos assigned brain then breast; breast: most common LFS cancer in women; first cancer: overall picture
Sex and brain
Men about twice the brain hazard (HR 2.15); sex-adjusted check did not change the mutation-type result
Functional-class grid question
Which cancer, given cancer: each class vs all other affected carriers
Five functional classes
Loss of function, dominant negative, gain of function, temperature-sensitive, partial function
Grid primary method
Logistic regression with cluster-robust standard errors on family
Grid sensitivity check
One row per family, Fisher's exact test
FDR method
Benjamini-Hochberg; star = q < 0.05
Grid pattern
LoF, DN, GoF lean to brain, sarcoma, leukemia, colorectal; TS and partial lean away; breast flat near 1
Link between grid and R337H
1,081 of 1,338 temperature-sensitive carriers are R337H: both analyses single out R337H
Germline check
Removed families flagged as blood-only (CHIP or blood cancer), not inherited
Software
Python, lifelines package (KaplanMeierFitter, CoxPHFitter)
Limit: one site
765 of 985 R337H carriers in the brain analysis come from one site
Future work 1
Correct for how families were found (Mendel, kin-cohort) with delayed entry and competing risk
Future work 2
Classify variants without a functional call; validate in carriers not found through cancer
So what
Mutation type could guide how closely carriers are watched, once corrected
TAD / PRD / DBD / TET / CTD
Transactivation, proline-rich, DNA-binding, tetramerization, C-terminal domains