Bioinformatics and Computational Biology Short Notes
One-page CSIR NET revision: databases, scoring matrices, BLAST, phylogenetics, annotation, variants, force fields, AlphaFold, drug design and systems biology.
- 10 min
On this page · 9 sections
The unit on one page. For explanations and worked problems, go back to the chapter: Bioinformatic Resources, Sequence Alignment, Phylogenetics, Comparative Genomics and Annotation, Molecular Modelling and Structure Prediction, Computer-Aided Drug Design and Systems Biology and Personalised Medicine.
Databases and Formats
| Fact | Detail |
|---|---|
| Primary vs secondary | Primary = submitted, redundant (GenBank, SRA, PDB, GEO); secondary = curated or derived (RefSeq, Swiss-Prot, Pfam, CATH) |
| INSDC | GenBank (NCBI, USA), ENA (EMBL-EBI, UK), DDBJ (NIG, Japan); same accessions; daily exchange |
| Versions | AB123456.2 = second version of the sequence |
| RefSeq prefixes | NC chromosome, NG gene region, NM mRNA, NR ncRNA, NP protein; X = predicted (XM, XP) |
| UniProtKB | Swiss-Prot reviewed, TrEMBL automatic; UniParc archive; UniRef100/90/50 clusters |
| Expression | GEO (GPL platform, GSM sample, GSE series, GDS dataset), ArrayExpress, Expression Atlas, GTEx; MIAME, MINSEQE |
| PDB | Experimental structures; wwPDB (RCSB, PDBe, PDBj, BMRB); EMDB maps; mmCIF standard; 4-character IDs |
| Structure quality | Resolution lower = better; R-free guards over-fitting; B-factor = mobility |
| AlphaFold DB | Predicted models, with pLDDT |
| Pattern/family | PROSITE (regex + profiles), PRINTS (fingerprints), BLOCKS (ungapped blocks → BLOSUM), Pfam (HMMs), SMART, CDD, InterPro (integrates) |
| PROSITE syntax | x any; [ST] either; {P} not P; x(2,4) range; < N-terminus, > C-terminus; N-glycosylation N-{P}-[ST]-{P} |
| GO | Molecular function, biological process, cellular component; DAG; IEA = electronic, unreviewed |
| Pathways / interactions | KEGG, Reactome, BioCyc / STRING, BioGRID, IntAct; SBML models in BioModels |
| Variants / disease | dbSNP (rs), dbVar, ClinVar, OMIM, gnomAD, COSMIC |
| Model organisms | FlyBase, WormBase, SGD, TAIR, MGI, ZFIN |
| Formats | FASTA (>), GenBank flat file, FASTQ (4 lines), SAM/BAM/CRAM, VCF, GFF3/GTF, BED 0-based half-open, PDB/mmCIF |
| Phred | Q = −10 log₁₀ P; Q20 = 1%, Q30 = 0.1%, Q40 = 0.01%; Phred+33 (‘I’ = Q40) |
| Coverage | C = LN/G; uncovered fraction e^(−C) |
Alignment and Database Search
| Fact | Detail |
|---|---|
| Homology | Common ancestry, yes or no; identity and similarity are percentages |
| Orthologue / paralogue | Speciation / duplication; xenologue = HGT; analogue = convergence |
| In-/out-paralogues | Duplication after / before the speciation considered |
| Dot plot | Diagonal = similarity; parallel diagonals = repeats; offsets = indels |
| Log-odds | s = (1/λ) ln(q_ab / p_a p_b) |
| PAM | Dayhoff; PAM1 = 1% change; extrapolated; higher PAM for distant |
| BLOSUM | Henikoff 1992; BLOCKS clustered at X%; lower BLOSUM for distant; BLOSUM62 default |
| BLOSUM62 diagonal | W–W 11, C–C 9 (rare residues score highest) |
| Affine gap | W = G_o + k G_e; blastp 11/1 |
| Needleman–Wunsch | Global; F(i,0) = ig; result bottom-right |
| Smith–Waterman | Local; floor 0; start at highest cell; stop at 0 |
| DP cost | ∝ m × n |
| FASTA | ktup 2 protein / 6 DNA; init1 → initn → opt (banded SW) |
| BLAST | Words (protein 3 classic, blastn 11, megablast 28); neighbourhood threshold T; extension; HSP; two-hit |
| BLAST programs | blastn, blastp, blastx (query translated), tblastn (database translated), tblastx (both) |
| E-value | E = Kmn e^(−λS); ∝ database size; lower = more significant |
| Bit score | S′ = (λS − ln K)/ln 2; E = mn 2^(−S′); database-independent |
| PSI-BLAST | Iterative PSSM; remote homologues; profile drift |
| MSA | Progressive (ClustalW, guide tree; “once a gap, always a gap”), iterative (MUSCLE, MAFFT), consistency (T-Coffee) |
| Clustal marks | * identical, : strong, . weak |
| Profiles | PSSM; profile HMM (match, insert, delete states) |
| Sequence logo | IC = log₂N − H; DNA max 2 bits |
Phylogenetics and Comparative Genomics
| Fact | Detail |
|---|---|
| Groups | Monophyletic (clade), paraphyletic, polyphyletic |
| Tree types | Cladogram (topology), phylogram (change), chronogram (time, ultrametric) |
| Rooting | Outgroup or midpoint |
| Tree counts | Unrooted (2n − 5)!!, rooted (2n − 3)!!; 4 taxa 3/15; 5 taxa 15/105 |
| JC69 | d = −¾ ln(1 − 4p/3); p = 0.30 → 0.383; saturation at 0.75 |
| K2P | Transitions vs transversions |
| Models | JC, K2P, F81, HKY, GTR, +Γ, +I; AIC/BIC |
| UPGMA | Clock; rooted ultrametric; average distances weighted by cluster size |
| NJ | Saitou and Nei 1987; no clock; unrooted |
| Parsimony | Fewest changes; informative sites; long-branch attraction |
| ML / Bayesian | P(data given tree and model) / posterior via MCMC |
| Bootstrap | Resample columns with replacement; % support (often ≥ 70% taken as good) |
| Clock | Pairwise d = 2rt |
| dN/dS | < 1 purifying, ≈ 1 neutral, > 1 positive |
| Comparative genomics | Synteny, collinearity, reciprocal best hits, COG/eggNOG, CNEs (phylogenetic footprinting), HGT signals |
| Pan-genome | Core + accessory + unique; open vs closed |
| C-value paradox | Genome size does not track complexity |
Annotation and Variation
| Fact | Detail |
|---|---|
| Structural vs functional | Where genes are vs what they do |
| Prokaryotic finders | ORF + Shine–Dalgarno; GeneMark, Glimmer (interpolated Markov models), Prodigal |
| Eukaryotic finders | GT–AG, Kozak, CpG islands, AATAAA; GENSCAN, AUGUSTUS (GHMMs); RNA-seq evidence; MAKER/BRAKER |
| Random ORFs | P(no stop in n codons) = (61/64)ⁿ; mean about 21 codons |
| ncRNA | tRNAscan-SE, Rfam |
| Sn / Sp (gene finding) | TP/(TP + FN) / TP/(TP + FP) |
| Variants | SNV (SNP if MAF ≥ 1%), indel, STR, CNV, SV; germline vs somatic |
| HGVS | c., g., p.; CFTR c.1521_1523delCTT, p.(Phe508del) |
| Pipeline | FASTQ → BWA/Bowtie2 → BAM → GATK/bcftools → VCF → VEP/SnpEff |
| Prediction | SIFT (≤ 0.05 deleterious), PolyPhen-2, CADD |
| LD | D = p_AB − p_A p_B; D′ = D/D_max; r² = D²/(p_A p_a p_B p_b) |
| GWAS | P < 5 × 10⁻⁸; Manhattan and QQ plots; PCA for stratification; OR = ad/bc; hit ≠ causal variant |
Modelling and Structure Prediction
| Fact | Detail |
|---|---|
| MM vs QM | MM cannot break bonds; QM/MM for reactions (Nobel 2013: Karplus, Levitt, Warshel) |
| Force field | Bonds + angles + torsions + impropers + LJ 12-6 + Coulomb |
| Lennard-Jones | Zero at σ; minimum −ε at 2^(1/6)σ ≈ 1.12σ |
| Coulomb | 1389 q₁q₂/(ε_r r) kJ/mol (r in Å); ε_r ≈ 4 interior, ≈ 80 water |
| Force fields | AMBER, CHARMM, OPLS-AA (all-atom); GROMOS (united-atom); MARTINI (coarse); MMFF94, GAFF (ligands) |
| Solvent | Explicit TIP3P, SPC/E; implicit PB, GB |
| Minimisers | Steepest descent (robust), conjugate gradient, Newton–Raphson/L-BFGS; local minimum only |
| Monte Carlo | Metropolis: accept uphill with e^(−ΔE/kT) |
| MD | F = −∇U; Verlet; 1–2 fs (SHAKE/LINCS); PBC; PME; Maxwell–Boltzmann start |
| Ensembles | NVE; NVT (thermostat); NPT (thermostat + barostat) |
| Analysis | RMSD (drift), RMSF (flexibility), Rg (compactness), PCA, MM-PBSA, FEP |
| Enhanced sampling | Replica exchange, metadynamics, umbrella sampling, steered MD |
| SCOP | Class → Fold → Superfamily (probable ancestry) → Family (clear relationship) |
| CATH | Class (1 α, 2 β, 3 α–β, 4 few SS) → Architecture → Topology → Homologous superfamily |
| Levinthal | 3^100 ≈ 5 × 10⁴⁷; folding funnel |
| 1D prediction | Chou–Fasman, GOR, PSIPRED (profiles); TM helices by hydropathy (~20 residues); SignalP |
| Homology modelling | Template → alignment → backbone → loops → side chains (rotamers) → refine → validate; MODELLER, SWISS-MODEL |
| Identity guide | > 50% high accuracy; 30–50% useful; < 30% twilight zone |
| Threading | Fold library + knowledge-based potentials; Phyre2, I-TASSER |
| Ab initio | Rosetta fragment assembly; coevolution contacts |
| AlphaFold2 | MSA + Evoformer + structure module; CASP14 (2020); Nobel 2024 (Hassabis, Jumper; Baker for design) |
| pLDDT | > 90 very high; 70–90 confident; 50–70 low; < 50 very low (often disorder) |
| PAE | Relative position confidence; high between domains = orientation uncertain |
| Successors | AlphaFold-Multimer, AlphaFold3 (ligands, nucleic acids), RoseTTAFold, ESMFold (single sequence) |
| Validation | Ramachandran (PROCHECK, MolProbity), ProSA, QMEAN; RMSD, GDT_TS, TM-score (> 0.5 same fold) |
Drug Design
| Fact | Detail |
|---|---|
| Pipeline | Target → hit → lead → optimisation → preclinical → Phase I–III |
| Chemical databases | PubChem (SID, CID, AID), NCI collections and the NCI-60 cell-line panel, ChEMBL, DrugBank, ZINC, BindingDB, PDBbind, CSD |
| Notations | SMILES, InChI/InChIKey, MOL/SDF, SMARTS (substructure patterns) |
| Affinity | \(K_d\) small = tight; \(\Delta G^{\circ} = RT\ln K_d\); 10× affinity ≈ 5.7 kJ mol⁻¹ (1.36 kcal mol⁻¹) |
| \(IC_{50}\) → \(K_i\) | Cheng–Prusoff \(K_i = IC_{50}/(1 + [S]/K_m)\) |
| Binding forces | H-bonds and ionic (enthalpic, desolvation cost), hydrophobic effect (entropic, dominant), van der Waals, conformational entropy loss |
| Binding models | Lock and key, induced fit, conformational selection |
| Ligand efficiency | \(LE = -\Delta G/N_{\text{heavy}}\) |
| SBDD | Needs the target structure; prepare → site → dock → rescore → assay → new structure |
| Docking | Search (systematic, genetic, Monte Carlo) + scoring (force-field, empirical, knowledge-based, ML); receptor usually rigid; redocking RMSD < 2 Å validates; scores ≠ free energies |
| Better energetics | MM-PBSA/GBSA, FEP, thermodynamic integration |
| Virtual screening | \(\text{EF}_{x\%}\) = (actives in top x%)/(expected); ROC-AUC; fragment-based, de novo, repurposing |
| LBDD | Similar property principle; descriptors 0D–3D (log P, TPSA, σ, \(E_s\)); Tanimoto \(T = c/(a + b - c)\) |
| Pharmacophore | 3D feature arrangement (donor, acceptor, hydrophobe, aromatic, ±ionisable); scaffold hopping; bioisosteres |
| QSAR | Hansch (log P, σ, \(E_s\)), Free–Wilson (additive); parabolic log P optimum |
| 3D-QSAR | CoMFA (steric + electrostatic fields + PLS), CoMSIA |
| QSAR validation | \(R^2\) (fit) vs \(q^2 > 0.5\) vs external \(R^2_{pred}\); y-randomisation; applicability domain |
| Rule of five | MW > 500, log P > 5, donors > 5, acceptors > 10; two or more violations flag poor oral absorption |
| Other filters | Veber (rotatable bonds ≤ 10, TPSA ≤ 140 Ų), rule of three (fragments), PAINS |
| ADMET | Solubility, permeability (Caco-2, PAMPA), bioavailability, protein binding, BBB and P-gp, CYP450 metabolism and interactions, \(t_{1/2} = 0.693V_d/CL\), toxicity (hERG, Ames) |
Systems Biology and Personalised Medicine
| Fact | Detail |
|---|---|
| Systems biology | Networks + models + high-throughput data; measure → model → predict → test |
| Emergent properties | Robustness, bistability, oscillation, ultrasensitivity, adaptation |
| Networks | PPI (STRING, BioGRID, IntAct), gene regulatory, metabolic (KEGG, MetaCyc), signalling (Reactome), co-expression (WGCNA), disease |
| Graph terms | Degree, hub, betweenness (bottleneck), clustering coefficient, module |
| Scale-free | \(P(k)\sim k^{-\gamma}\); robust to random loss, fragile to hub attack; small-world = short paths + clustering |
| Motifs | Negative autoregulation (fast, low noise); positive autoregulation (bistable); coherent FFL (delay/filter); incoherent FFL (pulse); SIM; DOR; delayed negative feedback (oscillation) |
| Model types | ODE, stochastic (Gillespie), Boolean (attractors), constraint-based (FBA), agent-based, PDE |
| Rate laws | Mass action; Michaelis–Menten; Hill \(n > 1\) ultrasensitive |
| Standards | SBML, BioModels, SBGN; COPASI, COBRA, CellDesigner |
| FBA | \(\mathbf{Sv} = 0\) + flux bounds + objective (biomass); no kinetic parameters, no time course; optima usually non-unique; FVA, pFBA, dynamic FBA |
| Data science traps | \(p \gg n\); multiple testing (Bonferroni = FWER, Benjamini–Hochberg = FDR/q); batch effects; FAIR data |
| Methods | PCA (linear) vs t-SNE/UMAP (visualisation only); clustering; differential expression; GO/pathway enrichment; LASSO, random forest, boosting, neural networks |
| ML discipline | Train/validation/held-out test, feature selection inside cross-validation, external cohort; class imbalance → use sensitivity, specificity, PR-AUC, ROC-AUC |
| Disease modelling | Disease modules, network pharmacology, driver vs passenger, SIR models (\(R_0\), herd immunity \(1 - 1/R_0\)), PBPK, digital twins |
| Digital health | EHR, telemedicine, wearables, decision support; FHIR/HL7, ICD, SNOMED CT, LOINC; privacy, bias, regulation, federated analysis |
| Personalised medicine | Pharmacogenomics (metaboliser status, HLA risk alleles, companion diagnostics), molecular stratification, targeted therapy, polygenic risk scores, liquid biopsy, ACMG/AMP interpretation |
| Test arithmetic | Sensitivity and specificity are prevalence-independent; PPV depends on prevalence; for rare conditions specificity dominates PPV |
| ELSI | Consent, incidental findings, right not to know, discrimination, data sovereignty, equity, ancestry bias in reference data |
Numbers to Have Ready
| Quantity | Value |
|---|---|
| Q30 error per base | 10⁻³ |
| Phred+33 ‘I’ | Q40 |
| e^(−3) | 0.050 |
| Bit score 50, m = 300, n = 10⁹ | E ≈ 2.7 × 10⁻⁴ |
| Extra raw score to cut E tenfold | ln 10/λ |
| JC d at p = 0.1 / 0.3 / 0.6 | 0.107 / 0.383 / 1.207 |
| (61/64)^100 | 0.008 |
| 2^(1/6) | 1.12 |
| 100 ns at 2 fs | 5 × 10⁷ steps |
| RT at 300 K | 2.49 kJ/mol |
| Metropolis, ΔE = 5 kJ/mol, 300 K | 0.13 |
| RT at 298 K | 2.48 kJ/mol; RT ln 10 = 5.7 kJ/mol |
| ΔG° at K_d = 1 nM | −51 kJ/mol |
| Tanimoto, a = 60, b = 80, c = 40 | 0.40 |
| PPV at 1% prevalence, 99% sensitivity, 95% specificity | 17% |
| Bonferroni α for 20,000 tests | 2.5 × 10⁻⁶ |
Traps in One Line Each
- “40% homologous” is wrong usage; say 40% identical.
- PAM250 and BLOSUM45 are for distant sequences.
- E-value rises with database size; bit score does not.
- Smith–Waterman traceback starts at the maximum cell.
- Human α-globin vs mouse β-globin: paralogues.
- UPGMA assumes a clock; NJ does not.
- Bootstrap % is support under resampling, not the probability of truth.
- GWAS lead SNP is usually a marker in LD, not the cause.
- Minimisation finds a local minimum, at 0 K, with no time axis.
- Same SCOP fold does not prove homology; same superfamily implies it.
- Low pLDDT often means disorder; inter-domain orientation needs PAE.
- AlphaFold2 models contain no ligands.
- A large K_d means weak binding, not tight.
- The rule of five flags trouble at two violations, and says nothing about potency or safety.
- Docking ranks poses well and affinities poorly; every hit still needs an assay.
- A high R² proves nothing; check q², an external test set and the applicability domain.
- FBA gives steady-state fluxes, not time courses, and its optimum is usually not unique.
- Scale-free networks are robust to random failure and fragile at hubs.
- For a rare disease, most positives from a good test are false: PPV follows prevalence.