Describe single-nucleotide polymorphisms (SNPs)

Overview

Single-nucleotide polymorphisms (SNPs) are the most common type of genetic variation in the human genome, occurring when a single nucleotide base differs between individuals at a specific position in DNA. Understanding SNPs is foundational to genomics, personalized medicine, and evolutionary biology.

---

Core Concept

---

The Molecular Mechanism

How SNPs Arise: Derivation from First Principles

Step 1: DNA Replication Errors

During S phase of the cell cycle, DNA polymerase copies ~3 billion base pairs. The polymerase has a natural error rate:

$\text{Error rate} = \frac{\text{Mismatched bases inserted}}{\text{Total bases replicated}} \approx 10^{-9} \text{ to } 10^{-10} \text{ per base per division}$

Why this step? Even with proofreading (3'→5' exonuclease activity), polymerase occasionally inserts the wrong nucleotide. For example, it might place a T opposite G instead of the correct C.

Step 2: Escape from Mismatch Repair

Most errors are caught by the mismatch repair (MMR) system, which scans for base-pair distortions. But if:

• MMR machinery is temporarily inefficient
• The error occurs in a heterochromatic region (less accessible)
• Random chance (~1 in 1000 errors escape)

Then the mismatch persists through the next round of replication.

Step 3: Fixation in the Population

If the mutation occurs in a germ cell (sperm/egg), it can be transmitted to offspring. Over generations, the probability that a single new copy eventually reaches 100% frequency (fixation) is:

\dfrac{1}{2N} & \text{if neutral (no selection)} \\[2ex] \dfrac{1-e^{-2s}}{1-e^{-4Ns}} & \text{if selected (selection coefficient } s) \end{cases}$$ Where $N$ = effective population size. **Why this formula?** Neutral mutations drift randomly; their fixation probability equals their initial frequency (1 copy in $2N$ chromosomes = $\frac{1}{2N}$). For the selected case, this is Kimura's diffusion result for a single new copy (initial frequency $p = 1/2N$). For a weakly favorable allele with small $s$, expanding $1-e^{-2s}\approx 2s$ gives the well-known approximation $P_{\text{fixation}}\approx 2s$ when $Ns \gg 1$. **Step 4: Becoming a "Polymorphism"** A new mutation becomes a SNP when its frequency crosses **1%** in the population. This takes: $$t_{\text{SNP threshold}} \approx \frac{\ln(0.01 \cdot 2N)}{s} \text{ generations (if selected)}$$ For neutral variants, drift alone requires thousands to millions of generations. --- ## Types of SNPs by Genomic Location >[!formula] SNP Classification by Functional Impact >SNPs are first sorted by **where** they fall relative to genes. Because most of the genome is non-coding, the vast majority of SNPs are intronic or intergenic. Within the **coding** subset, we then further classify by protein effect. > >**(a) Genome-wide location of ALL SNPs (sums to ~100%):** > >| **Location** | **Approx. share of all SNPs** | **Functional Effect** | >|----------|---------------|----------------------| >| **Intergenic** | ~45% | Often neutral; some affect long-range regulation | >| **Intronic** | ~45% | Usually silent; rare splice-site effects | >| **Regulatory** (promoter/enhancer/UTR) | ~8% | Alters gene expression levels | >| **Coding (exonic)** | ~1–2% | May change protein (see below) | > >**(b) Within the CODING subset only, by protein effect:** > >| **Coding type** | **Approx. share of coding SNPs** | **Functional Effect** | >|----------|---------------|----------------------| >| **Synonymous** | ~50% | No amino acid change (wobble position) | >| **Non-synonymous (missense)** | ~48% | Amino acid substitution | >| **Nonsense** | ~2–4% | Premature stop codon | > >**Why two tables?** Table (a) percentages describe *all* SNPs and sum to ~100%. Table (b) percentages are conditional — they only describe the small (~1–2%) coding slice, and *those* sum to ~100% among coding SNPs. Mixing the two scales is the classic error that makes numbers appear to exceed 100%. ### Example 1: Synonymous SNP (Silent) **Gene:** hypothetical coding region, a cysteine codon **Reference codon:** ...**TGC**... → Cys (Cysteine) **SNP variant:** ...**TGT**... → Cys (Cysteine) **Why this step?** The genetic code is degenerate: both **TGC** and **TGT** code for cysteine (they differ only at the wobble 3rd position C↔T). The SNP changes the DNA but not the encoded amino acid. **Impact:** Generally neutral, but can affect mRNA stability or translation speed (codon usage bias). --- ### Example 2: Non-synonymous SNP (Missense) **Gene:** *HBB* (β-globin), codon 6 **Reference (HbA):** ...GTG **GAG** GTG... → Glu (Glutamic acid, charged) **SNP variant (HbS):** ...GTG **GTG** GTG... → Val (Valine, hydrophobic) **Why this step?** A single nucleotide change of the middle base (A→T) turns the codon **GAG** (Glu) into **GTG** (Val), replacing a hydrophilic amino acid with a hydrophobic one. **Molecular consequence:** 1. **Hydrophobic patch forms** on hemoglobin surface 2. Under low oxygen, HbS molecules **polymerize** into rigid fibers 3. Red blood cells **sickle** (deform into crescent shape) $$\text{Sickling tendency} \propto e^{-\Delta G_{\text{polymerization}}/RT}$$ Where $\Delta G_{\text{polymerization}}$ becomes negative (favorable) when Val-Val hydrophobic interactions outweigh the entropy cost of alignment. **Why this matters?** This single SNP causes sickle cell disease (homozygotes) but confers malaria resistance (heterozygotes)—a classic case of **balancing selection**. --- ### Example 3: Regulatory SNP **Gene:** *LCT* (lactase) promoter region, position -13910 (in intron 13 of *MCM6*) **Reference (most mammals):** ...C... → **Lactase shuts off after weaning** **SNP variant (T allele):** ...T... → **Lactase persists into adulthood** **Mechanism:** 1. The T allele creates a binding site for the transcription factor **Oct-1** 2. Oct-1 keeps the *LCT* promoter **active** throughout life 3. Lactose (milk sugar) remains digestible in adults **Population genetics:** $$\text{Frequency of T allele} = \begin{cases} >90\% & \text{Northern Europe (dairy farming cultures)} \\ <10\% & \text{East Asia, Sub-Saharan Africa} \end{cases}$$ **Why this distribution?** The T allele spread rapidly (~10,000 years ago) in populations that domesticated cattle, providing a nutritional advantage. This is **positive selection** in action. --- ## SNP Density and Distribution ### Calculating SNP Frequency **Total human genome size:** $L = 3.2 \times 10^9$ base pairs **Average SNPs between two individuals:** $N_{\text{SNP}} \approx 4-5 \times 10^6$ **SNP density:** $$\rho_{\text{SNP}} = \frac{N_{\text{SNP}}}{L} = \frac{4 \times 10^6}{3.2 \times 10^9} = 1.25 \times 10^{-3} = \frac{1}{800 \text{ bp}}$$ **Why this matters?** On average, every ~800 base pairs contains one SNP between any two unrelated humans. This means: - We're **99.9% identical** at the DNA level - That 0.1% difference accounts for all observable human variation ### Non-Random Distribution SNPs are NOT uniformly distributed: 1. **Recombination hotspots:** High SNP density (crossover shuffles variants) 2. **Coding regions:** Lower SNP density (purifying selection removes harmful variants) 3. **CpG islands:** High mutation rate (methylated cytosines spontaneously deaminate to thymine) $$\mu_{\text{CpG}} \approx 10 \times \mu_{\text{other}}$$ **Why?** Methylated cytosine (5-methylcytosine) is chemically unstable and converts to thymine, creating C→T transitions. --- ## Detecting and Using SNPs ### Genotyping Methods >[!example] SNP Detection Workflow >**1. DNA Microarray (SNP Chip)** >- Probe sequences complementary to reference and variant alleles >- Sample DNA hybridizes to matching probes >- Fluorescent labels distinguish alleles >- **Throughput:** 500,000 - 5 million SNPs per sample >**2. Sequencing-Based Detection** >- Whole-genome sequencing reveals ALL SNPs >- Bioinformatics pipeline: > ``` > Raw reads → Alignment → Variant calling → Filtering (quality score >30, depth >10×) > ``` >- **Advantage:** Discovers novel SNPs, not just known ones --- ### Example 4: Genome-Wide Association Study (GWAS) **Question:** Which SNPs are associated with Type 2 Diabetes? **Method:** 1. Genotype 500,000 SNPs in 10,000 diabetics and 10,000 controls 2. For each SNP, calculate association strength: $$\chi^2 = \frac{(O - E)^2}{E}$$ Where $O$ = observed allele frequency in cases, $E$ = expected frequency from controls. **Result interpretation:** $$P\text{-value} = P(\chi^2 > \text{observed}|\text{null hypothesis})$$ Significance threshold (Bonferroni correction): $$P_{\text{threshold}} = \frac{0.05}{500{,}000} = 10^{-7}$$ **Why this threshold?** Testing 500,000 SNPs inflates false-positive risk. Bonferroni divides the alpha level by the number of tests to maintain family-wise error rate at 5%. **Why this step?** SNPs near genes like *TCF7L2* (transcription factor 7-like 2) show $P < 10^{-10}$, indicating strong association. The SNP itself may not be causal—it's often in **linkage disequilibrium** with the true causal variant. --- ## Linkage Disequilibrium (LD) >[!formula] LD Explained >**Definition:** Non-random association of alleles at two loci. >If SNP A (alleles A/a) and SNP B (alleles B/b) are in LD: >$$D = P(AB) - P(A)P(B)$$ >**$D = 0$:** Alleles assort independently (linkage equilibrium) >**$D \neq 0$:** Alleles are inherited together more (or less) often than chance > >**Normalized measure (r²):** > >$$r^2 = \frac{D^2}{P(A)P(a)P(B)P(b)}$$ > >- **$r^2 = 1$:** Perfect LD (two SNPs always inherited together) >- **$r^2 = 0$:** No LD (independent) **Why LD matters for SNPs:** In GWAS, a "hit" SNP may just be a proxy for the real causal SNP nearby. You need fine-mapping or sequencing to find the actual functional variant. **Decay of LD with distance:** $$r^2(d) \approx \frac{1}{1 + 4Nrc}$$ Where $c$ = recombination rate per bp per generation, $d$ = distance between SNPs. **Why this step?** Recombination breaks up haplotypes over generations. Distant SNPs eventually reach equilibrium. --- ## Common Mistakes >[!mistake] Mistake 1: "SNPs are rare mutations" >**Why it feels right:** The word "polymorphism" sounds exotic, and we learn about disease mutations being rare. > >**The fix:** By definition, SNPs have **≥1% frequency**—they're COMMON. Rare variants (<1%) are just called "mutations" or "rare variants," not SNPs. The 1% cutoff is arbitrary but standard. > >**Steel-man:** This confusion arises because disease-causing SNPs (like the sickle cell SNP) get disproportionate attention, even though most SNPs are neutral or have tiny effects. >[!mistake] Mistake 2: "All SNPs in coding regions change the protein" >**Why it feels right:** We focus on dramatic examples like HbS, so it seems like SNPs always alter function. > >**The fix:** Due to **codon degeneracy**, roughly half of coding SNPs are synonymous (silent). Even non-synonymous SNPs often involve conservative substitutions (e.g., Leu→Ile, both hydrophobic) with minimal impact. > >**Test yourself:** Given a random coding SNP, what's the rough probability it's synonymous? > >$$P(\text{synonymous}) \approx \frac{\text{\# synonymous mutation possibilities}}{\text{total possible single-base changes}} \approx \tfrac{1}{4}\text{–}\tfrac{1}{2}$$ >[!mistake] Mistake 3: "Percentages of SNP types all sum to 100% on one scale" >**Why it feels right:** Tables list percentages, so you assume they belong to one pie. > >**The fix:** Location percentages (intergenic + intronic + regulatory + coding ≈ 100%) are on the **whole-genome** scale. Synonymous/non-synonymous/nonsense percentages are **conditional on being coding** (a ~1–2% slice). Mixing scales makes the numbers appear to exceed 100%. >[!mistake] Mistake 4: "Tag SNPs detect all variation in a region" >**Why it feels right:** GWAS uses "tag SNPs" to represent haplotype blocks efficiently, so it seems like they capture everything. > >**The fix:** Tag SNPs only capture **common variation** (MAF >5%). Rare variants, structural variants, and de novo mutations are missed. This is the "missing heritability" problem—GWAS SNPs explain only ~10-20% of trait heritability. --- ## Applications of SNPs 1. **Personalized Medicine:** Pharmacogenomics uses SNPs in drug-metabolizing enzymes (e.g., *CYP2D6*) to predict drug response 2. **Ancestry Testing:** SNP patterns reveal population origins and migration histories 3. **Evolution Studies:** SNP divergence between species estimates divergence time: $$T_{\text{divergence}} = \frac{D_{\text{SNP}}}{2\mu}$$ Where $D_{\text{SNP}}$ = fraction of differing SNPs, $\mu$ = mutation rate per site per year 4. **Crop Breeding:** SNP markers enable marker-assisted selection for yield, disease resistance --- ## Active Recall >[!recall]- Feynman Explanation (ELI12) >Imagine DNA is like a recipe book for building you. Every person's recipe book has 3 billion letters. Now, if you compare your book with your friend's book, almost all the letters are the same—like 99.9% identical! But in a few spots, maybe one letter is different. Instead of "Add Salt," your book might say "Add Sult" (silly example, but you get it). > >These one-letter differences are called SNPs (snips). Most of them don't do anything important—like a typo that doesn't change the meaning. But some SNPs matter a lot. For example, there's a SNP that decides whether you can digest milk as an adult. One letter change in your DNA turns on a gene that breaks down milk sugar. If you don't have that SNP, drinking milk gives you a stomachache! > >Scientists love studying SNPs because they explain why people look different, why some people get certain diseases, and even where your ancestors came from. It's like a genetic fingerprint, but way more detailed than an actual fingerprint. >[!mnemonic] SNP Memory Aids >**"SNP = Single Nucleotide Polymorphism"** >- **S**ingle: ONE base change only >- **N**ucleotide: A, T, G, or C >- **P**olymorphism: "Poly" = many forms; common (>1%) in population >**For types:** **"SNIR"** = **S**ynonymous, **N**on-synonymous, **I**ntronic, **R**egulatory >**For LD:** "LD Links Loci" = Linkage Disequilibrium means alleles at different loci are inherited together --- ## Connections - [[Genetic Variation and Mutations]] — SNPs are one type of genetic variant - [[DNA Replication Fidelity]] — Replication errors generate SNPs - [[Population Genetics and Hardy-Weinberg]] — SNP allele frequencies follow HW equilibrium - [[Genome-Wide Association Studies (GWAS)]] — Uses SNPs to map disease genes - [[Linkage Disequilibrium and Haplotypes]] — Explains why nearby SNPs correlate - [[Molecular Evolution and Neutral Theory]] — Most SNPs are neutral (Kimura's theory) - [[Pharmacogenomics]] — SNPs predict drug metabolism - [[Sickle Cell Disease]] — Classic non-synonymous SNP example --- #flashcards/biology What is the formal definition of a SNP? ::: A single nucleotide variation at a specific genomic position that occurs in ≥1% of the population, involving substitution of one base (A, T, G, or C) only. What is the average SNP density in the human genome? ::: Approximately 1 SNP per 800 base pairs, or ~4-5 million SNPs between two unrelated individuals (0.1% of the genome). Why does the genetic code's degeneracy matter for SNPs? ::: About half of coding SNPs are synonymous (do not change the amino acid) because multiple codons encode the same amino acid, especially at the wobble (3rd) position (e.g., TGC and TGT both = Cys). What causes the HbS (sickle cell) SNP's molecular effect? ::: A single A→T mutation at the middle base of codon 6 of β-globin changes GAG (Glu, hydrophilic) to GTG (Val, hydrophobic), creating a hydrophobic patch that causes hemoglobin polymerization under low oxygen. What is the frequency threshold that distinguishes SNPs from rare mutations? ::: 1% (or minor allele frequency MAF ≥ 0.01). Variants below this are typically called rare mutations, not SNPs. How does linkage disequilibrium (LD) affect GWAS interpretation? ::: A GWAS hit SNP may not be the causal variant—it's often in LD with (inherited alongside) the true functional SNP nearby. Fine-mapping or sequencing is needed to identify the actual causal variant. What is the Bonferroni-corrected significance threshold for a GWAS testing 500,000 SNPs? ::: P < 10^-7 (calculated as 0.05 / 500,000 to control family-wise error rate for multiple testing). Why are CpG sites hotspots for SNPs? ::: Methylated cytosines at CpG dinucleotides spontaneously deaminate to thymine at ~10× the normal mutation rate, causing frequent C→T transitions. What is the fixation probability of a single new neutral mutation in a diploid population of size N? ::: 1/(2N), because a neutral allele's fixation probability equals its initial frequency (1 copy among 2N chromosomes). What is Kimura's fixation probability for a single new selected allele? ::: P = (1 − e^(−2s)) / (1 − e^(−4Ns)); for small favorable s with Ns ≫ 1 this approximates to ≈ 2s. Give an example of a regulatory SNP and its effect. ::: The -13910 C→T SNP upstream of the lactase gene (LCT) creates an Oct-1 transcription factor binding site, maintaining lactase expression into adulthood (lactase persistence). What percentage of the human genome differs between two unrelated individuals due to SNPs? ::: Approximately 0.1% (we are 99.9% identical), accounting for ~4-5 million SNP differences out of 3.2 billion base pairs. Why do coding regions have lower SNP density than intergenic regions? ::: Purifying (negative) selection removes deleterious SNPs that alter protein function, while intergenic SNPs are more often neutral and persist. ## 🖼️ Concept Map ```mermaid flowchart TD REP[DNA replication] -->|polymerase error| ERR[Mismatched base] ERR -->|caught by| MMR[Mismatch repair] ERR -->|escapes MMR| MUT[Persistent mutation] MUT -->|in germ cell| GERM[Heritable variant] GERM -->|drift or selection| FIX[Fixation in population] FIX -->|reaches ≥1% frequency| SNP[Single-nucleotide polymorphism] RARE[Rare mutation] -->|below 1% threshold| SNP SNP -->|single base substitution| SUB[One nucleotide swap C to T] SNP -->|explains| TRAIT[Trait variation e.g. caffeine metabolism] SNP -->|used in| MED[Personalized medicine] SNP -->|tracks| DIS[Disease susceptibility] SNP -->|studied in| EVO[Evolutionary biology] ``` ## 🔊 Hinglish (regional understanding) > [!intuition]- Hinglish mein samjho > Dekho, DNA ko tum ek 3 billion letters wali instruction manual samjho jo har insaan mein almost same hoti hai. Par kuch specific positions par ek single letter alag hota hai - jaise ek jagah tumhare paas 'C' hai aur tumhare dost ke paas 'T'. Yeh single-letter changes hi **SNPs** (snip bolte hain) kehlate hain, aur yahi human genome mein sabse common genetic variation hai. Important baat yeh hai ki isko SNP tabhi bolte hain jab yeh variation population ke kam se kam 1% logon mein present ho - warna woh ek rare mutation hai. Yeh ek single nucleotide (A, T, G, ya C) ka substitution hota hai, insertion ya deletion nahi. > > Ab yeh matter kyun karta hai? Yeh SNPs tumhare genetic "fingerprints" hain - yehi explain karte hain ki kyun koi banda coffee jaldi digest kar leta hai, kyun kisi ko lactose intolerance hai, aur kyun kisi ko koi particular disease hone ka zyada chance hai. Iski wajah se hi **personalized medicine** possible hoti hai, jahan doctor tumhare genes dekh ke batata hai ki kaunsi dawai tumpe best kaam karegi. Mechanism simple hai: DNA replication ke time polymerase kabhi kabhi galat nucleotide daal deta hai (error rate bahut kam hoti hai), aur agar mismatch repair system usko catch nahi kar paaya, toh woh error permanent ho jaata hai. > > Ek key point yaad rakhna - agar yeh mutation germ cell (sperm ya egg) mein hua, tabhi yeh next generation mein transmit hota hai aur dheere-dheere population mein spread hota hai. Neutral mutation ka fixation probability bas $\frac{1}{2N}$ hoti hai (jahan N population size hai), matlab bina kisi selection advantage ke bhi random drift se yeh fix ho sakta hai. Toh short mein, SNPs woh "permanent typos" hain jo genetic diversity create karte hain aur genomics ki backbone hain - inko samajh gaye toh personalized medicine aur evolution dono ka foundation clear ho jaayega. ![[audio/6.1.08-Describe-single-nucleotide-polymorphisms-(SNPs).mp3]]