Describe non-coding DNA and the ENCODE findings

What is Non-coding DNA?

Why So Much Non-coding DNA?

Historical context: When the protein-coding paradigm dominated molecular biology (1960s-1990s), scientists focused on the ~1.5% of DNA that encodes proteins. The rest was dismissed as "junk DNA"—evolutionary baggage with no function.

The puzzle: If natural selection removes useless material, why hasn't evolution eliminated 98% of our DNA over millions of years? Three possibilities emerged:

1. It's truly junk (neutral drift maintains it)
2. It has structural roles (spacing, chromosome architecture)
3. It's functionally critical (but not for making proteins)

The ENCODE Project: A Paradigm Shift

Key ENCODE Findings (2012 Main Phase)

Finding 1: Most of the genome is transcribed

• ~75-80% of the genome is transcribed into RNA at some point in development or in specific cell types
• Why this matters: Transcription is energetically expensive. If a sequence is transcribed, there's selective pressure to maintain it—suggesting function
• The mechanism: RNA polymerase II doesn't just transcribe protein-coding genes; it produces thousands of long non-coding RNAs (lncRNAs), enhancer RNAs (eRNAs), and antisense transcripts

Finding 2: Regulatory elements are pervasive

• Identified ~400,000 enhancer regions and ~70,000 promoter regions
• ~80% of the genome participates in at least one biochemical event (transcription factor binding, histone modification, chromatin remodeling)
• Why this matters: Each protein-coding gene needs multiple regulatory elements to control when, where, and how much it's expressed. A typical gene might have 5-10 enhancers spread across hundreds of kilobases

Finding 3: Functional elements are evolutionarily constrained

• Many non-coding regions show sequence conservation across species (slower mutation rate than neutral expectation)
• Disease-associated SNPs (from GWAS studies) overwhelmingly fall in non-coding regulatory regions (~93%), not in protein-coding sequences
• Why this matters: Mutations in regulatory DNA can cause disease by changing gene expression levels, even when the protein itself is normal

The Biochemical vs. Functional Debate

Derivation: Why Conservation Implies Function

Let's derive the relationship between mutation rate and selective constraint quantitatively.

Setup: Consider a non-coding DNA sequence. Let:

• $\mu$ = neutral mutation rate (mutations per base pair per generation)
• $s$ = selection coefficient (fitness cost of a deleterious mutation: $0\leq s \leq 1$; here $s>0$ means deleterious, reducing fitness by $s$)
• $N_e$ = effective population size

Step 1: Neutral expectation For truly junk DNA ($s = 0$), mutations accumulate at the neutral rate. After $t$ generations of divergence between two lineages:

$k_{\text{neutral}} = 2\mu t$

Why this step? Under neutrality, the substitution rate equals the mutation rate. The rate of new neutral mutations entering the population is $2N_e\mu$ (per generation, diploid), and each has fixation probability $1/(2N_e)$, so the substitution rate is $2N_e\mu \times \tfrac{1}{2N_e} = \mu$. The factor of 2 accounts for both diverging lineages.

Step 2: Fixation probability of a deleterious mutation For a new mutation with selection coefficient $s$ (deleterious, so we use $-s$), Kimura's diffusion result gives the fixation probability:

$P_{\text{fix}}(s) = \frac{1 - e^{-2s}}{1 - e^{-4N_e s}}\cdot \frac{1}{2N_e}$

Why this step? A new mutation starts at frequency $p = 1/(2N_e)$. Kimura's general formula for fixation probability of an allele starting at frequency $p$ with selection coefficient $s$ is $u(p) = \frac{1 - e^{-4N_e s\,p}}{1 - e^{-4N_e s}}.$ Substituting $p = 1/(2N_e)$ gives $4N_e s\,p = 2s$, yielding the expression above. Note the sign convention: for a deleterious allele you replace $s \to -s$, giving $P_{\text{fix}} = \dfrac{1 - e^{2s}}{1 - e^{4N_e s}}\cdot\dfrac{1}{2N_e}$, which is small for large $N_e s$.

Sanity checks (Why this step?):

• Neutral limit ($s \to 0$): L'Hôpital gives $P_{\text{fix}} \to \dfrac{1}{2N_e}$ — exactly the neutral fixation probability. ✓
• Strongly deleterious ($N_e s \gg 1$, deleterious): $P_{\text{fix}} \to 0$ — bad mutations are purged. ✓

Step 3: Constrained substitution rate The substitution rate for the constrained sequence is (rate of new mutations) × (fixation probability):

$k_{\text{constrained}} = 2t\cdot\big(2N_e\mu\big)\cdot P_{\text{fix}}(s)$

Step 4: Conservation ratio Define the conservation score $C$ as the ratio of observed to neutral divergence:

$C = \frac{k_{\text{constrained}}}{k_{\text{neutral}}} = \frac{2t\,(2N_e\mu)\,P_{\text{fix}}(s)}{2\mu t} = 2N_e\,P_{\text{fix}}(s)$

Substituting the deleterious Kimura result $P_{\text{fix}} = \dfrac{1 - e^{2s}}{1 - e^{4N_e s}}\cdot\dfrac{1}{2N_e}$:

$\boxed{\,C = \frac{1 - e^{2s}}{1 - e^{4N_e s}}\,}$

Interpreting the box (Why this matters):

• For neutral DNA ($s \to 0$): $C \to 1$ (evolves at the neutral rate — no conservation).
• For deleterious/constrained DNA ($N_e s \gg 1$): the denominator dominates and $C \to 0$ (strong conservation — substitutions are suppressed).
• So a low $C$ (e.g., $C = 0.1$, evolving at 10% of neutral rate) implies the sequence is under selective constraint, i.e., functional.

Types of Functional Non-coding DNA

1. Regulatory Elements

2. Non-coding RNAs

3. Repetitive DNA

Historically dismissed as "junk," but now recognized for functions:

• Transposable elements (~45% of human genome, combining LINEs ~21%, SINEs ~13%, LTR/retroviral elements ~8%, and DNA transposons ~3%): Source of regulatory innovation. When they insert near genes, they can bring new regulatory sequences. Example: a significant fraction of human promoters and enhancers contain transposon-derived sequences
• Satellite DNA: Structural role at centromeres (chromosome segregation) and telomeres (chromosome end protection)
• Tandem repeats: Variable number tandem repeats (VNTRs) affect gene expression. Example: A 44 bp repeat in the SLC6A4 promoter (serotonin transporter) affects transcription efficiency and is linked to anxiety/depression susceptibility

Clinical Relevance: Disease SNPs in Non-coding DNA

Computational Prediction of Functional Elements

ENCODE developed machine learning models to predict functional non-coding DNA:

Features used:

• DNase I hypersensitivity (open chromatin, accessible to proteins)
• Histone modifications (e.g., H3K4me3 at promoters, H3K27ac at active enhancers)
• Transcription factor ChIP-seq (direct protein-DNA binding)
• Evolutionary conservation (PhastCons, PhyloP scores)

Output: ChromHMM (Chromatin Hidden Markov Model) segments the genome into chromatin states:

• Active promoter (H3K4me3, H3K27ac, DNase I)
• Strong enhancer (H3K4me1, H3K27ac, DNase I)
• Polycomb-repressed (H3K27me3)
• Heterochromatin (H3K9me3)
• Quiescent (no marks)

Validation: When ENCODE predictions are tested with reporter assays (insert predicted enhancer → measure expression), a large fraction show enhancer activity.

Implications for Medicine and Biotechnology

1. Genome editing precision: CRISPR must consider non-coding DNA. Editing a gene for therapy might inadvertently delete an enhancer for a different gene nearby.

2. Personalized medicine: Most disease risk is in regulatory DNA. Understanding which non-coding variants affect drug response or disease susceptibility will be crucial.

3. Synthetic biology: Designing synthetic regulatory circuits requires understanding endogenous regulatory logic. ENCODE provides a parts list for building genetic switches.

4. Evolutionary biology: Non-coding DNA is the substrate for morphological evolution. Changes in body plan (e.g., loss of hindlimbs in snakes) often result from regulatory mutations, not protein-coding mutations.

Connections

• Gene Regulation and Expression Control - How non-coding DNA controls when and where genes turn on
• Chromatin Structure and Histone Modifications - Epigenetic marks that ENCODE maps
• RNA Processing and Splicing - Introns are non-coding DNA removed during processing
• Evolutionary Conservation and Phylogenetics - How we detect functional elements by comparing species
• GWAS and Complex Trait Genetics - Disease SNPs mostly in regulatory DNA
• CRISPR and Genome Editing - Must consider non-coding DNA for precision editing
• Long Non-coding RNAs - XIST and other functional RNA genes
• Enhancer-Promoter Interactions - How distant regulatory elements control genes
• Dosage Compensation and X-inactivation - XIST-mediated silencing
• Transposable Elements and Genome Evolution - Repetitive DNA as source of regulatory innovation

#flashcards/biology

Concept Map

Hinglish (regional understanding)