1 of 22

DNA methylation clock

Saket Choudhary

saketc@iitb.ac.in

Computational Multi-omics of Ageing

DH 603

Lecture 02 || Friday, 7th March 2025

2 of 22

History of ageing

3 of 22

History of ageing: Gilgamesh

  • 4000 year old Epic of Gligamesh → from Mesopotamia
  • The titular hero, king of the city of Uruk, befriends a wild warrior Enkidu
  • Enkidu unfortunately dies motivating Gilgamesh to conquer death
  • Gilgamesh consults Utnapishtim who offers him two routes to immortality:
    • 1: Escape death if you stay awake for a week straigh
      • But he falls asleep and fails
    • 2: Consume a herb found underwater
      • Gilgamesh successfully obtains it and plans to test it by first feeding it to the older people of Uruk
      • But a snake slithers away with the herb

4 of 22

History of ageing: China

  • First emperor of unified China
  • Having won many battles, conquered territories, he wanted to win the elixir of life
  • Sent missionaries to find the elixir with a punishment of execution if they fail to do so
  • Would ultimately die before constructing his mausoleum possibly due to mercury poisoning from one of the potions he took to prolong his life

5 of 22

The story of culturing immortality

  • Alexis Carrel was a famous (Nobel laureate) french surgeon famous for pioneering techniques to connect severed blood vessels
  • With the goal of keeping tissues alive, he began a series of experiments at Rockefeller Institute to keep a culture of cells alive from a tissue outside the human body, indefinitely
  • Took cells from heart of chicken embryo, kept steadily supplying them with nutrients → the culture could be maintained for years → big breakthrough
  • 1921 issue of Scientific American reads, “Perhaps the day is not far away when most of us may reasonably anticipate a hundred years of life. And if a hundred, why not a thousand?”

But Carrel’s finding could not be replicated in other labs for years!

6 of 22

The Hayflick Limit

  • Overturned the “central dogma” that all vertebrate cells grown in culture are immortal
  • Hayflick and Moorhead performed an experiment using human fibroblasts which showed that fibroblasts doubled a finite number of times, after which the cells stopped dividing and entered what Hayflick termed the phase III phenomenon
  • Simple experiment: Mix equal numbers of normal human male fibroblasts that had divided many times with female fibroblasts that had divided only a few time.
  • When the male ‘control’ culture stopped dividing, the mixed culture was examined and only female cells were found → old cells ‘remembered’ they were old, even when surrounded by young cells

7 of 22

A brief timeline

8 of 22

Is human lifespan limited?

9 of 22

Supercentenarians database

10 of 22

Is there a limit to human lifespan?

  • A hard limit can exist only if a programmed process leading to death exists
  • The human lifespan can be modeled using statistical distribution using variations on the generalized extreme value distribution (GEV) → statistical model of the probability distribution of the oldest age a person can survive to
  • After fitting the statistical model, if the shape parameter is negative (the orange curve), the density function has support (nonzero probability) from –∞ to ω → there is no chance for anyone to live past age ω (a hard limit to lifespan)
  • When the shape parameter is 0 (the blue curve), the density function has support over the interval (–∞,+∞); survival to any age is possible, although trending very close to zero for higher ages (a soft limit to lifespan)

11 of 22

Methylation clocks

12 of 22

DNA-sequences can undergo modifications

Fi

  • A family of enzymes (DNA methyltransferase, DNMT) catalyze the transfer of methyl group to the 5th carbon of cytosine to form 5-methylcytosine (5mC)
  • Primarily occurs in the CpG (C followed by a G connected by a phospodiester bond
  • DNMT3a and DNMT3b are de-novo methyl transferase → transfer methyl group to naked DNA
  • DNMT1 is a maintenance methyltransferase

5mC

13 of 22

Methylation happens primarily at CpG sites

Fi

  • CpG sites (C followed by G) occur with high frequency in the genome
  • Cytosines in CpG dinucleotides can be methylated to form 5-methylcytosines → Mammals have 70-80% CpGs methylated
  • Methylated CpG can change the expression of the nearby gene

14 of 22

Variation of methylation status across contexts

Fi

  • Methylation quantified as a percentage of ‘reads’ originating from a loci that are methylated (details later)
  • Mammalian ‘methylomes’ have pervasive methylation (80% of CpG sites are methylated in humans)
  • Organism like bees have sporadic methylation
  • Organisms like drosophila have almost no methylation → methylation is not a requirement for development or fate specification
  • Absence of conservation of methylation across organisms hints that it is not a

15 of 22

Methylation arrays capture CpG sites across genome

Fi

  • Methylation arrays capture 27k (Illumina 27k array) 450k (Illumina 450k array) or 850k (“EPIC”) CpG sites across the genome spanning ~20k genes
  • Bisuflite treatment converts unmethylated C to U
  • Bead contains sequences that hybridize to sequences across the genome (containing the CpG dicnuleotide)
  • Beta values = Methylated C (Methylated C + Unmethylated U)

Constant offset

16 of 22

Methylation clock - the science

Steve Horvath

17 of 22

Some definitions

Chronological age Age of an organism as measured by physical time, for example, calendar years from birth.

Biological age A correlative measure of chronological age that is more informative than chronological age of the current and future health status of an individual. An example of a directly quantifiable biological age is mortality risk.

18 of 22

Chronological age is captured by CpG methylation

7.5k samples from 27k or 450k arrays→ 353 informative CpGs

19 of 22

Works across species

20 of 22

Ever expanding table of clocks

21 of 22

Elastic Net

  • First term encourages sparse solution to coefficients
  • Second term encourages highly correlated features to be averaged

Linear regression

Under constraints on beta

22 of 22

22

Questions?