Describe genome annotation — Hinglish notes

6.1.6 · Biology › Genomics

Core Concept

Genome annotation ek systematic process hai jo genome sequence mein saare functional elements ko identify aur label karta hai — jismein genes, regulatory elements, repeat sequences, aur doosre biologically meaningful features shamil hain.

HUM YE KYUN CHAHIYE? Ek raw genome sequence mein zero context hota hai — yeh aise hai jaise kisi encyclopedia ke saare letters random order mein daal diye gaye hon. Annotation sequence data ko biological knowledge mein transform karta hai.

YE KAISE KAAM KARTA HAI? Computational prediction algorithms ke through, jo experimental validation ke saath combine hote hain.

The Annotation Pipeline

1. Structural Annotation (Genes dhundhna)

Formula Gene Prediction Approaches

Ab initio (de novo) prediction:

Gene structure ke statistical models use karta hai (Hidden Markov Models, neural networks)
Signals dhundhta hai: start codons (ATG), stop codons (TAA, TAG, TGA), splice sites (GT-AG), promoter motifs
Probability ki derivation: Ek sequence $S$ ke liye, hum chahte hain $P (gene ∣ S)$ $P (gene ∣ S) = \frac{P ( S ∣ gene ) \cdot P ( gene )}{P ( S )}$

Jahan:

$P (S ∣ gene)$ = likelihood ki sequence known genes jaisi lagti hai (training data se seekhi gayi)
$P (gene)$ = prior probability (roughly genome mein gene density)
$P (S)$ = normalizing constant

YE KYUN KAAM KARTA HAI: Genes mein non-random patterns hote hain — codon usage bias, length distributions, conserved splice sites. Model ye patterns verified genes se seekhta hai.

Homology-based prediction:

BLAST ya similar alignment use karke doosre organisms ke known genes se compare karo
Agar human insulin gene ke saath 80% sequence identity milti hai, to shayad ye ek insulin gene hai $Similarity Score = \frac{Number of matches}{Alignment length} \times 100%$

Evidence-based prediction:

Actual experimental data use karo: RNA-seq (dikhata hai kya transcribe hota hai), proteomics (dikhata hai kya translate hota hai)
Yeh ground truth hai — agar tum mRNA sequence karo aur wo ek genomic region se align ho, to wo region definitely transcribed hai

2. Functional Annotation (Genes KYA KARTE HAIN?)

Worked example Example 1: Ek Bacterial Gene ko Annotate Karna

Scenario: Tumne E. coli mein 1,200 bp ka ek open reading frame (ORF) identify kiya hai.

Step 1 - Structural:

Position 1000 par start codon ATG mila (ORF ka pehla base)
Positions 2197–2199 par stop codon TAA mila (ORF ke aakhri teen bases)
ORF length = 2199 − 1000 + 1 = 1200 bp → 400 codons → 399 amino acids (stop codon koi residue code nahi karta)

YE STEP KYUN? Hum confirm kar rahe hain ki iske paas ek gene ki basic structure hai: ek start codon, ek contiguous in-frame reading region, aur ek proper 3-base stop codon (TAA/TAG/TGA). Coordinates consistent hone chahiye — start position + length − 1 = stop end position.

Step 2 - Homology Search:

NCBI database ke against BLAST
Top hit: E. coli lactose permease (lacY), 95% identity YE STEP KYUN? Sequence similarity functional similarity imply karta hai (homologs usually function share karte hain).

Step 3 - Domain Search:

Pfam search "Major Facilitator Superfamily" domain dhundhta hai
Ye domain membrane transport proteins mein milta hai

YE STEP KYUN? Protein domains functional units hain — known domains dhundhne se hume molecular function pata chalta hai.

Step 4 - Pathway Analysis:

KEGG database dikhata hai ki ye gene "Galactose metabolism" pathway mein hai
lacZ (beta-galactosidase) aur lacA (transacetylase) ke saath interact karta hai YE STEP KYUN? Function context se aata hai — pathway placement jaanne se biological meaning milta hai.

Final Annotation:

Gene: lacY_homolog
Product: Lactose permease
Function: Transmembrane transport of lactose
Pathway: Galactose metabolism (KEGG map00052)
GO Terms: GO:0015153 (galactose transmembrane transporter activity)

Worked example Example 2: Alternative Splicing wale Human Gene ko Annotate Karna

Scenario: Tumhe RNA-seq evidence ke saath ek 50,000 bp genomic region mila hai.

Step 1 - Exon Detection:

RNA-seq reads 5 distinct regions (exons) mein map hote dikhate hain
Splice junction reads confirm karte hain: Exon1-Exon2, Exon1-Exon3, Exon2-Exon4, Exon3-Exon4, Exon4-Exon5

YE KYUN MATTER KARTA HAI? Ye gene alternative splicing ke through multiple isoforms produce karta hai. Step 2 - Isoform Reconstruction:

Isoform A: E1-E2-E4-E5 (E3 skip karta hai)
Isoform B: E1-E3-E4-E5 (E2 skip karta hai) Step 3 - Functional Prediction:
Isoform A: 350 amino acids, kinase domain contain karta hai → probably ek signaling enzyme hai
Isoform B: 280 amino acids, kinase domain truncated → probably ek dominant-negative regulator hai

ALAG FUNCTIONS KYUN? Alternative splicing often functional domains remove ya add karta hai, alag ya opposing activities wale proteins create karta hai.

Annotation Challenge: Dono isoforms ko unke tissue-specific expression patterns ke saath record karna zaroori hai.

Annotation Quality Metrics

Formula Sensitivity, Specificity aur Precision

Gene prediction tools ko TP, FP, FN, TN ke chaar counts wale confusion matrix se evaluate kiya jaata hai. Teen key metrics:

$Sensitivity (Recall) = \frac{T P}{T P + F N}$

$Specificity = \frac{T N}{T N + F P}$

$Precision (Positive Predictive Value) = \frac{T P}{T P + F P}$

Jahan:

TP = True Positives (correctly identified genes)
FP = False Positives (predicted genes jo real nahi hain)
FN = False Negatives (real genes jo miss ho gaye)
TN = True Negatives (non-gene regions jo correctly non-gene call kiye gaye)

TEEN METRICS KYUN?

Sensitivity poochta hai: saare real genes mein se, hum kitne pakad sake?
Specificity poochta hai: saare non-gene regions mein se, hum ne kitne correctly chod diye?
Precision poochta hai: jo kuch bhi humne genes kaha, unme se kitne actually genes hain?

Dhyan raho — ek common confusion: TP/(TP+FP) precision hai, specificity NAHI. Specificity true negatives use karta hai: TN/(TN+FP). Gene finding mein, precision often report hoti hai kyunki true negatives (saare non-coding positions) enormous hain aur define karna mushkil hai.

SENSITIVITY–PRECISION TRADE-OFF KYUN?

High sensitivity (saare genes pakadna) → zyada false positives → lower precision
High precision (sirf confident calls) → rare ya unusual genes miss ho jaate hain → lower sensitivity

Typical tools well-studied organisms par ~90% sensitivity aur ~85% precision achieve karte hain.

Types of Annotation Challenges

Challenge 1: Non-coding RNA genes

KYA: Genes jo protein ki jagah functional RNA produce karte hain (tRNA, rRNA, miRNA, lncRNA) MUSHKIL KYUN: Koi open reading frame nahi, koi start/stop codons nahi — secondary structure ya conservation par rely karna padta hai

Challenge 2: Pseudogenes

KYA: Gene-jaise sequences jo function kho chuke hain (often stop codons ya frameshifts hote hain) IMPORTANT KYUN: Real genes se confuse ho sakte hain; pseudogenes ko samajhna evolutionary history reveal karta hai

Challenge 3: Overlapping genes

KYA: Ek gene doosre ke andar encoded (opposite strand ya different reading frame) MUSHKIL KYUN: Standard gene-finders non-overlapping genes assume karte hain; special detection chahiye

Common mistake Common Annotation Errors

Galti 1: "Har ORF ek gene hai" Ye sahi kyun lagta hai: Open reading frames (start → stop bina interruptions ke) genes jaisi lagti hain.

Ye galat kyun hai: Random DNA mein, stop codons frequently aate hain — stop codons ke beech mean run sirf 63 bp (≈ 21 codons) hota hai, kyunki 64 codons mein se 3 stops hain. Isliye chhote ORFs constantly hote hain, lekin genuinely lambe ORFs (>300 bp) random sequence mein statistically rare hain. Ek chhota ORF essentially chance se expected hai aur khud mein almost kuch bhi matlab nahi rakhta.

Fix: Additional evidence chahiye — known genes se homology, RNA-seq support, species across conservation, aur random expectation se lambi length. ORF ek candidate hai, confirmation nahi.

Galti 2: "High BLAST similarity = identical function"

Ye sahi kyun lagta hai: Agar do proteins 80% identical hain, to shayad woh same kaam karte hain. Ye galat kyun hai: Chhote sequence differences function dramatically change kar sakte hain, aur bahut low identity bhi ek shared fold ke andar chhup sakti hai. Example: rhodopsin (light sensor) aur olfactory receptors dono class-A GPCR family mein hain aur sirf 20–25% identity share karte hain, phir bhi bilkul alag ligands aur functions hain. Similar fold ≠ identical function.

Fix: Similarity ko starting hypothesis ki tarah use karo, phir check karo:

Kya key catalytic residues conserved hain?
Kya domain architecture match karta hai?
Kya expression pattern similar hai?

Galti 3: "Koi known homolog nahi = gene nahi" Ye sahi kyun lagta hai: Agar BLAST kuch nahi dhundhta, shayad ye junk DNA hai.

Ye galat kyun hai: Har gene family ka ek "pehla member" tha jiske koi homologs nahi the. Rapidly evolving genes (jaise immune system ya toxin genes) mein often koi detectable homologs nahi hote. Fix: Evidence ki multiple lines use karo — RNA-seq, proteomics, synteny ki evolutionary conservation (gene order), purifying selection signatures.

The Human Genome Annotation Saga

Automated vs Manual Annotation

Aspect	Automated	Manual (Curation)
Speed	Fast (millions of genes/day)	Slow (hours per gene)
Consistency	High	Variable
Accuracy	~85-90%	~95-99%
Cost	Low	High (expert time)
Best for	First-pass annotation	Model organisms, clinical genes

Reality: Genome-wide coverage ke liye automated use karo, important genes (disease-related, drug targets) ke liye manual curation karo.

Recall Ek 12 saal ke bachche ko explain karo

Socho tumhe abhi ek bahut bada LEGO instruction manual mila hai, lekin sab kuch code mein hai — sirf 1s aur 0s. Genome annotation aise hai jaise ek computer (aur smart log) us manual se guzar rahe hain aur har part label kar rahe hain: "Ye numbers matlab hai car banao," "Ye matlab hai ghar banao," "Ye color scheme ke instructions hain."

Lekin yahan tricky part hai: kabhi kabhi instructions overlap karte hain, ya backwards likhe hote hain, ya kuch instructions doosre instructions ki broken copies hain. Annotation team ko figure out karna hota hai ki kya real instruction hai, kya galti hai, aur har instruction actually kya banata hai.

Jab scientists ek genome annotate karna finish karte hain, tab woh jaante hain: "Is organism mein lagbhag 20,000 alag protein-building instruction sets (genes) hain, aur hum confident hain ki unme se zyaadatar kya banate hain." Annotation ke bina, genome useless data hai. Annotation ke saath, ye life ka blueprint hai.

Connections

6.1.01-DNA-sequencing-methods - Annotation ke liye pehle high-quality sequence chahiye
6.1.05-Comparative-genomics - Cross-species comparison annotation accuracy improve karta hai
Gene-expression-profiling - RNA-seq transcribed regions ka direct evidence provide karta hai
Protein-structure-prediction - Functional annotation 3D structure jaanne se benefit karta hai
BLAST-and-sequence-alignment - Homology-based annotation ka core tool
Hidden-Markov-Models - Ab initio gene prediction ka statistical framework

#flashcards/biology

Genome annotation kya hai? :: Ek genome sequence mein functional elements ko identify aur label karne ka systematic process, jismein genes, regulatory elements, aur repeat sequences shamil hain.

Genome annotation ke do main types kya hain?

Structural annotation (genes, exons, introns ki locations/structures identify karna) aur functional annotation (identified genes ko biological functions assign karna).

Ab initio gene prediction kya hai?

Ek computational method jo statistical models use karke sequence mein hi patterns ke basis par genes predict karta hai, bina known genes se comparison kiye.

Homology-based annotation kya hai?

BLAST jaise tools use karke genome sequence ko doosre organisms ke known genes ke databases se compare karke genes identify karna.

Hum har open reading frame (ORF) par ek real gene ke roop mein trust kyun nahi kar sakte?

Random DNA mein stop codons ke beech mean run sirf ~63 bp hota hai, isliye chhote ORFs constantly chance se aate hain; sirf lambe ORFs (>300 bp) jo supporting evidence ke saath hain (RNA-seq, homology, conservation) likely real hain.

Gene prediction mein specificity aur precision mein kya fark hai?

Specificity = TN/(TN+FP), non-gene regions ka fraction jo correctly exclude kiye gaye. Precision (PPV) = TP/(TP+FP), predicted genes ka fraction jo real hain. Ye same NAHI hain.

Gene prediction mein sensitivity (recall) kya hai?

Sensitivity = TP/(TP+FN), real genes ka fraction jo correctly identify kiye gaye.

Non-coding RNA genes annotate karna mushkil kyun hai?

Inme open reading frames aur translation signals nahi hote, isliye annotation secondary structure predictions ya evolutionary conservation patterns par rely karna padta hai.

Pseudogene kya hai?

Ek gene-jaise sequence jisne mutations (stop codons, frameshifts) ke through function kho diya hai, often ek real gene jaisa lagta hai lekin functional product produce nahi karta.

2001 se ab tak human gene count time ke saath kyun ghata hai?

Better algorithms aur experimental validation ne reveal kiya ki bahut saari initial predictions actually pseudogenes, annotation errors, ya true protein-coding genes ki jagah non-coding elements thi.

Kaun se types of experimental evidence gene annotations validate karte hain?

RNA-seq (transcription dikhata hai), proteomics (translation dikhata hai), ChIP-seq (regulatory binding dikhata hai), aur species across evolutionary conservation.