Describe next-generation sequencing (NGS)

6.1.4 · Biology › Genomics

Overview

Next-Generation Sequencing (NGS) ne genomics mein revolution la diya — iska kaam hai massively parallel DNA sequencing karna — matlab ek saath millions of fragments padhna, jabki Sanger method ek-ek karke padhta tha. Is parallelism ki wajah se genome ki cost ~$100M(2001)seghatkar<$1000 (2020s) ho gayi, aur projects jo saalon mein hote the, woh ab dinon mein ho jaate hain.

[!intuition] Core Idea Kya Hai

Socho jaise ek kitaab padhna ho agar:

1. Use kaat diya jaye laakhon overlapping tukdon mein
2. Saare tukde ek saath padhe jaayein (ek page ke baad doosra nahi)
3. Original text ko dobara banaaya jaaye overlaps dhundh ke

Sanger sequencing = ek magnifying glass se ek ek sentence padhna. NGS = saare sentences ek saath wall par project karna aur unki photo khींchna.

Parallel kyun? DNA polymerase (jo "reader" hai) ~1000 bp/s ki speed se kaam karta hai. Millions of clusters par parallelize karne se yeh biological speed limit bypass ho jaati hai — jaise ek photocopier ke bajaye millions of photocopiers hona.

[!definition] NGS Kya Hai?

Yeh sequencing technologies ki ek family hai jo ek saath millions se billions of short reads (50–300 bp) generate karti hai, is tarah se:

1. Library preparation: DNA ko fragment karna + adapters add karna
2. Clonal amplification: Identical fragments ke clusters banana
3. Sequencing-by-synthesis: Real-time mein saare clusters mein nucleotide incorporation detect karna
4. Bioinformatics: Reads ko reference se align karna ya de novo assemble karna

Key principle: Spatial separation + optical detection. Har DNA cluster ek chip/bead par ek unique location par hota hai; cameras har imaging cycle mein millions of sites par fluorescence signals record karte hain.

[!formula] Sequencing Depth aur Coverage

1. Coverage Depth

$C=\frac{N\cdot L}{G}$ Jahaan:

• $C$ = average coverage (kitne reads har genomic position par overlap karte hain)
• $N$ = total number of reads
• $L$ = read length (bp)
• $G$ = genome size (bp)

First principles se derivation:

• Total bases sequenced = $N\cdot L$ (reads × length/read)
• Yeh bases $G$ size ke genome mein randomly distribute hoti hain
• Kisi bhi position ke padhne ki average times = $\frac{\text{total bases}}{G}$

Example: Human genome ($G=3\times 10^9$ bp), 100M reads ($N=10^8$), 150 bp reads ($L=150$): $C=\frac{10^8\cdot 150}{3\times 10^9}=\frac{1.5\times 10^{10}}{3\times 10^9}=5\text{×}$ Yeh kyun matter karta hai: Variant calling ke liye 30× recommend kiya jaata hai (ensure karta hai ki ~99.9% genome ≥5 times cover ho, jo true SNPs ko errors se alag karne ke liye zaroori hai).

2. Lander-Waterman Coverage Probability

Random sequencing mein, probability ki ek base ≥1 baar cover ho: $P(\text{covered})=1-e^{-C}$

Derivation:

• Assume karo reads randomly place hue hain (Poisson process)
• Probability ki koi specific base ek read se hit nahi hoti = $1-\frac{L}{G}$
• $N$ independent reads ke liye: $P(\text{not covered})={\left(1-\frac{L}{G}\right)}^N$
• Bade $G$ ke liye, ${\left(1-\frac{L}{G}\right)}^N\approx e^{-NL/G}=e^{-C}$
• Toh $P(\text{covered})=1-e^{-C}$

Example: 5× coverage par: $P(\text{covered})=1-e^{-5}=1-0.067=0.993\text{ (99.3\%)}$ 30× par: $P=1-e^{-30}\approx 1.000$ (>99.999%)

Yeh step kyun? Exponential decay dikhata hai ki returns diminishing hote hain: 5× se 10× jaane par sirf 0.7% zyada coverage milti hai, lekin cost double ho jaati hai. Yahi math experimental design guide karta hai.

[!example] Illumina Sequencing-by-Synthesis (Sabse Common NGS)

Step-by-step with WHY:

1. Library Preparation

• Kya: DNA ko ~300–500 bp mein fragment karo, dono ends par adapters (short synthetic oligos) ligate karo.
• Fragment kyun? Clustered conditions mein polymerase sirf ~500 bp tak high fidelity se extend kar sakta hai.
• Adapters kyun? Primers ke liye binding sites provide karte hain (step 3) aur flow cell se attachment enable karte hain.

2. Cluster Generation (Bridge Amplification)

• Kya: Fragments ko flow cell surface par laga do jo adapters ke complementary oligos se coated hai. Har fragment adjacent oligo ki taraf jhukta hai ("bridge" banaata hai), prime aur extend hota hai → har cycle mein copies double hoti hain. Sirf ~10–12 cycles mein cluster ~1000 identical copies par plateau kar jaata hai (physical crowding ~1 µm patch par aage doubling ko rokta hai; agar tum 30 cycles chalao toh ~$2^{30}\approx 10^9$ molecules milte, jo tiny cluster area mein nahin sama sakte).
• Amplify kyun? Ek molecule ka fluorescence detect karne ke liye bahut dim hota hai. 1000 copies signal ko 1000× boost karte hain.
• Bridge kyun? Physical separation spatial resolution maintain karta hai — har cluster ~1 µm apart hota hai, camera se resolve hota hai.

3. Sequencing Cycles

• Kya: Fluorescently labeled reversible terminator nucleotides (A/T/G/C har ek ke paas unique dye + blocking group hoti hai) add karo. Saare clusters ke upar flow karo → har cluster mein ek base incorporate hoti hai. Poore flow cell ki image lo → har position par color record karo. Terminator + dye cleave karo → repeat.
• Terminator kyun? Ensure karta hai ki sirf ek base har cycle mein add ho (warna polymerase unevenly aage bhaag jaata).
• Reversible kyun? Agle cycle ke liye 3'-OH group restore karna hota hai.
• Example: Cycle 1: Cluster A green glows (C) → Cycle 2: red (T) → Cycle 3: green (C) → read = CTC...

4. Base Calling

• Kya: Software intensity images (4 channels: A=green, C=blue, G=yellow, T=red) ko quality scores ke saath FASTQ files mein convert karta hai.
• Quality scores kyun? Har base call ki error probability hoti hai (~0.1%–1%); Phred scores ($Q=-10{\log }_{10}{P}_{\text{error}}$) downstream analysis ko inform karte hain (e.g., $Q<30$ wale bases filter karo).

5. Alignment/Assembly

• Kya: Reads ko reference genome se map karo (BWA-MEM) ya de novo assemble karo (SPAdes).
• Reference kyun? Resequencing (variant calling) ke liye, known structure leverage karta hai. Novel genomes ke liye, overlaps assemble karne padte hain.

[!example] Ion Torrent (Semiconductor Sequencing)

Illumina se difference: Fluorescence nahi, pH change detect karta hai.

Kaise: Nucleotide incorporation se H⁺ release hota hai: $\text{DNA-3’-OH}+\text{dNTP}\stackrel{\text{polymerase}}{\to }\text{DNA-3’-dNMP}+\text{PPi}+{\text{H}}^{+}$

• Kya: Ek time par ek dNTP type flow karo (e.g., sirf dATP). Base pairing rule: A, T se pair karta hai. Toh jahan bhi template mein T ho, polymerase A add karta hai → H⁺ release → pH drop. Sensor #H⁺ ke proportional voltage shift detect karta hai.
• Yeh kyun kaam karta hai? Koi labels nahi chahiye → faster, cheaper reagents. Lekin homopolymers (AAAAA) mushkil hote hain: 5 A's 5× H⁺ release karte hain, lekin signal perfectly linear nahi hota.

Example: Template 3'→5' padhna = TTACG (woh strand jo copy ho raha hai). Complementary bases add hote hain:

• dATP flow karo: template positions jahan T hai → 2 A's TT par incorporate hote hain → large signal (2 H⁺).
• dTTP flow karo: template A → 1 T incorporate → small signal.
• dGTP flow karo: template C → 1 G incorporate → small signal.
• dCTP flow karo: template G → 1 C incorporate → small signal.

Yeh step kyun? Signal tab aata hai jab flowed nucleotide template base se match kare (base-pair kare). dATP signal precisely wahaan produce karta hai jahan template T hai, kyunki A–T pair banta hai.

[!mistake] Common Misconceptions

Mistake 1: "NGS reads Sanger se longer hote hain"

Kyun sahi lagta hai: "Next-gen" zyada advanced lagta hai, aur Sanger reads ~800 bp hote hain. Kyun galat hai: Illumina reads 50–300 bp hote hain (shorter!) kyunki parallelization length ko throughput se trade karta hai. "Generation" throughput ke baare mein hai, read length ke nahi. Fix: Long-read tech (PacBio, Nanopore: 10–100 kb) structural variants ya assembly ke liye use karo. Illumina ko SNPs/indels ke liye use karo jahan short reads kaafi hain.

Mistake 2: "30× coverage ka matlab hai har base 30 baar padha gaya"

Kyun sahi lagta hai: "30×" literally 30 reads per position jaisa lagta hai. Kyun galat hai: Coverage ek average hai. Random sampling se variance create hoti hai — kuch regions ko 50× milta hai, doosron ko 10×. Mean $\lambda =30$ wale Poisson model ke under, sirf lagbhag 1–2% positions 20× se neeche jaati hain (aur sirf ~2–3% 22× se neeche). Baat sahi hai — coverage uneven hoti hai — lekin shortfall utna nahi hota jitna log sochte hain. Fix: Gaps identify karne ke liye uneven coverage plots use karo; variant calls ke liye minimum depth require karo (e.g., ≥10×).

Mistake 3: "Zyada coverage always accuracy improve karta hai"

Kyun sahi lagta hai: Zyada data = better. Kyun galat hai: Sequencing errors kuch contexts mein systematic hote hain (PCR duplicates, GC bias). 100× par, tum wohi errors resample kar rahe ho, correct nahi kar rahe. Fix: Depth ko orthogonal QC ke saath pair karo (e.g., SNPs phase karne ke liye long reads, structural variants ke liye optical mapping). ~50× ke baad zyaataar applications mein diminishing returns shuru ho jaate hain.

[!recall]- Ek 12-Saal ke Bachche ko Explain Karo

Socho tumhare paas 1000-page ki ek kitaab hai, lekin kisine saare pages nikaal ke unhe confetti mein kaat diya. Tumhara kaam: pata lagao kitaab mein kya tha.

Purana tarika (Sanger): Ek confetti ka tukda uthao, microscope se dhyan se padho, likh lo. Agli tukda uthao. Yeh forever lagta hai!

NGS tarika: Millions of confetti tukde ek giant table par failao. Ek photo lo jahan har tukda alag color mein glow karta hai apne pehle letter ke hisaab se. Ab tum saare tukdon ka pehla letter ek saath jaante ho. Colors change karo doosra letter dikhane ke liye → aur photo lo. 150 baar repeat karo → tumne har tukde se 150 letters simultaneously padh liye.

Phir computer se overlaps dhundho: "Yeh tukda '...the cat' pe khatam hota hai aur woh tukda 'cat sat..' se shuru hota hai → yeh connect hote hain!" Yeh laakhon baar karo → kitaab reconstruct ho gayi.

Yeh faster kyun hai? Har tukda 10 seconds mein ek-ek padhne ke bajaye (10M tukde = 3 saal!), sab ko parallel mein padho (150 photos = 1 din).

[!mnemonic] NGS = "Naive Granny Sequences"

• Naive: DNA randomly fragment karta hai (specific regions nahi chunता)
• Granny: Parallel mein kaam karta hai (jaise ek saath 1000 scarves bunna)
• Sequences: By synthesis (DNA build karta hai padhte waqt)

Ya: "Lots of Clusters In Parallel" → Library, Cluster, Image, Process

Connections

• Sanger Sequencing - first-generation method; NGS ise parallelize karta hai
• DNA Polymerase - sequencing-by-synthesis mein use hone wala enzyme
• Bioinformatics Alignment Algorithms - BWA, Bowtie NGS reads map karte hain
• Genomic Variants (SNPs, Indels) - NGS inhe population scale par identify karta hai
• CRISPR - NGS edits aur off-target effects verify karta hai
• Metagenomics - NGS mixed microbial communities sequencing enable karta hai
• Cancer Genomics - tumor-normal NGS somatic mutations dhundta hai
• Personalized Medicine - NGS whole-genome sequencing affordable banata hai

#flashcards/biology

NGS ka full form kya hai aur Sanger sequencing ke muqable mein iska key advantage kya hai? :: Next-Generation Sequencing; yeh ek saath millions of DNA fragments simultaneously parallel mein sequence karta hai, jabki Sanger ek-ek karta tha, genome ki cost ~$100Mse<$1000 tak kam kar deta hai.

Concept Map