Biochem Lab Techniques: Gel Filtration, SDS-PAGE, X-ray Crystallography, and AlphaFold

Gel Filtration vs SDS-PAGE and Protein Quaternary Structure

• Context: The speaker contrasts purification/analysis methods in biochem labs and connects them to how we infer protein structure and composition.
• Core message: Different techniques reveal different aspects of a protein. Gel filtration (size-exclusion) preserves native quaternary structure; SDS-PAGE denatures proteins and reveals subunit composition, not the intact oligomer.

Gel filtration (size-exclusion chromatography)

• Proteins are not denatured during gel filtration; you observe native, folded oligomers.
• Purpose: estimate size/oligomeric state in native condition.
• Notation: In gel filtration you do not boil the sample in SDS or apply reducing conditions; thus the quaternary structure can be intact.

SDS-PAGE (Sodium dodecyl sulfate polyacrylamide gel electrophoresis)

• Denatures proteins, coating them with charge; runs separate by molecular weight of subunits under denaturing conditions.
• Typical visualization: dye binding (Coomassie blue dye is the common stain; transcript says "Kamasi blue"—likely a mispronunciation; standard is Coomassie blue).
• Visualization: blue bands appear where proteins bind the dye; migration distance correlates with molecular weight.
• Example interpretation: If you boil a protein complex and run SDS-PAGE, you observe individual subunits rather than the intact complex.

Interpreting a purification example (illustrative workflow from the transcript)

• Homogenate -> lysosome/plasma membrane fractions -> cytosol (the liquid portion) containing thousands of proteins; note that the actual SDS-PAGE result would look more complex before purification.
• He describes a change in the visualization when showing a purified product: histidine tag (hist tag) discussions appear—intended to illustrate that tagging and purification steps alter which proteins dominate in the sample.
• The example focuses on quaternary structure and how to deduce it from SDS-PAGE results.

Quaternary structure concepts

• Quaternary structure definitions:
  • Monomer: single subunit
  • Dimer, Trimer, Tetramer: oligomeric states with 2, 3, or 4 subunits
  • Homo: subunits identical
  • Hetero: subunits different
• The classic example in the transcript: a protein with total molecular weight 100 (target) made of subunits each weighing 25.
  • If you see a single band at 25 kDa on SDS-PAGE, you infer a tetramer of identical subunits in the native complex (HOMO tetramer).
  • If your observed bands sum to the total molecular weight, you can deduce the likely quaternary structure.
  • Calculation shown: total mass = 100; subunit mass = 25; number of subunits = 100 / 25 = 4; thus a HOMO tetramer.
• If SDS-PAGE shows more than one band, you infer multiple subunits or heteromeric composition.
  • Example scenario: two bands at 25 kDa and 75 kDa could indicate a tetramer composed of two 25-kDa subunits and one 50-kDa subunit, or other combinations that sum to 100; the details depend on the exact subunit masses.
  • If three bands appear, you may have a hetero-oligomer with three different subunits (or a more complex composition); the teacher notes this could be a puzzle-type test question.
• Important reminder given in the talk:
  • If there is only one band on SDS-PAGE, it suggests all subunits are identical (a homo-oligomer).
  • If there are multiple bands, you must consider both the individual subunit masses and how they could sum to the native protein’s total mass.

Test-taking context (study strategy and expectations)

• The instructor emphasizes there is no grade curve in the course.
• He discourages last-minute cramming; he wants students to consistently study rather than rely on a final-minute effort.
• He warns that exams can be challenging and puzzle-like, requiring you to deduce structures from given data (e.g., subunit composition from SDS-PAGE bands).
• Realistically, you should expect test questions that require you to reason about oligomeric state from band patterns and total molecular weight, not just memorize facts.

Overview of X-ray crystallography (big topic toward the exam)

• Why it matters: X-ray crystallography has been foundational in biochemistry; much of textbook knowledge comes from this technique.
• The lab approach described:
  • Crystallize a protein; in class you see the diffraction pattern and get a ribbon diagram; in reality, the process involves going from diffraction data to a 3D model.
  • The instructor plans to demystify the jump from diffraction pattern to a molecular ribbon model.
• Key steps in practice: 1) Crystallization of biological macromolecules (and nucleic acids). Crucial concept: purity and crystallinity. 2) Crystallization method demonstration: sitting-drop technique.
  • Sitting drop plate with wells; you place ~2 μL of protein solution in the well and seal with tape to create a closed micro-environment.
  • Evaporation increases protein concentration in the drop; as concentration increases, crystals may form as the solution becomes supersaturated.
  • The packing-tape trick is a humorous description of creating a closed environment for crystallization; other methods exist too.
    3) Crystal lattice and unit cell concepts:
  • A crystal is a repeating array of unit cells; the lattice is the order, and the actual crystal is the visible object.
  • Visualize a unit cell as the smallest repeating block in 3D space that builds the entire crystal.
    4) X-ray interaction with crystals:
  • X-rays interact with electrons; the diffraction pattern arises from scattering by electrons in each atom.
  • The intensity of a diffraction spot is related to the amplitude of the scattered waves, which in turn depends on electron density and arrangement.
  • The location of diffraction spots is determined by the crystal lattice (reciprocal lattice), i.e., the geometry of the unit cell.
    5) Fourier transform concept (Fourier’s theorem) as the core mathematical bridge:
  • Diffraction data (spots) are related to the electron density via a Fourier transform.
  • In crystals, the electron density ρ(r) can be reconstructed by summing a series of waves (Fourier synthesis) corresponding to the observed diffraction spots.
  • The general idea: complex periodic function (crystal structure) can be represented as a sum of sine waves; the diffraction spots provide coefficients (amplitude and phase) for those waves.
    6) Scattering factors: amplitude and phase concepts:
  • Each diffraction spot has an amplitude (intensity) and a phase; both are needed to reconstruct electron density.
  • In practice, phases are not directly observed and must be inferred (phasing problem).
    7) Reciprocal space and real space:
  • The reciprocal lattice points correspond to possible reflections; the measured pattern encodes the crystal’s symmetry and lattice parameters.
  • The distance between spots encodes unit-cell dimensions; the pattern’s symmetry reflects crystal symmetry.
    8) From electron density to a model:
  • After obtaining electron density maps from Fourier synthesis, researchers build a model by placing atoms (often starting with amino-acid residues) into density.
  • The example mentions lysine and glutamic acid placement to illustrate how amino-acid side chains fit into electron density.
    9) Refinement and validation:
  • Refinement: adjust the model to best fit the observed diffraction data by minimizing discrepancies between observed and calculated structure factors.
  • The R factor (R-work) is a key statistic:
    • Definition (typical form): $R = \frac{\sum<em>{hkl} |F</em>{obs} - F<em>{calc}|}{\sum</em>{hkl} |F_{obs}|}$
    • Interpretation: lower R indicates a better fit; perfectly modeled structure would give R = 0, but in practice acceptable values are often around 0.20 or less (20% or lower).
    • The instructor notes: 0 is ideal but rarely achieved; 0.20 or lower is considered acceptable in real-world structures.
  • Validation geometry: ensuring peptide bonds are planar due to resonance; Ramachandran plot checks on φ-ψ angles to ensure they fall in allowed regions; outliers lead to automatic rejection by validation.
  • The so-called X-ray cops in New Jersey joke alludes to the peer-review/validation process for crystal structures before publication.
    10) From diffraction to a reliable structure:
  • The process uses Fourier synthesis to move from observed diffraction data to electron density maps, and then to a final atomic model, followed by refinement and validation.

Culmination: AlphaFold and AI in structural biology

• AlphaFold 3 released in May (latest in the transcript): AI-based protein structure prediction.
• Comparison to experimental data:
  • AlphaFold predictions can be extremely accurate and sometimes rival experimentally determined structures for many proteins.
  • The transcript emphasizes that AlphaFold is powerful but not a substitute for experimental data; predictions are informed by, and compared to, existing experimental structures in databases.
• Ethical/practical considerations:
  • AI predictions should complement, not replace, experimental validation.
  • Experimental data (e.g., X-ray crystallography, cryo-EM) remains critical for confirmation, especially where structural details affect function or drug design.

Key conceptual takeaways and how they connect

• The same protein can appear differently depending on the method:
  • Gel filtration reveals native oligomeric state; SDS-PAGE reveals subunit composition once denatured.
• Quaternary structure depends on subunit masses and stoichiometry; you can infer this from mass data and band patterns if you know subunit sizes and total complex mass.
• X-ray crystallography: a mathematically grounded, highly quantitative method that translates a crystal’s periodic arrangement into a 3D electron-density map and a final atomic model via Fourier synthesis and refinement.
• The Fourier connection is central: observed diffraction patterns (reciprocal space) encode information about real-space electron density and atomic coordinates.
• Validation and realism: geometry (planarity of peptide bonds), φ/ψ angle regions (Ramachandran), and R-factors are essential checks; unrealistic models are rejected.
• AI advances (AlphaFold) are transformative but do not eliminate the need for experimental data; they provide strong predictions that can guide experiments and interpretation.

Important equations and concepts to remember

• Quaternary structure inference (conceptual): if total protein mass is M and subunit mass is m, then number of subunits n is
  $n = \frac{M}{m}.$
  Example: if M = 100 and m = 25, then $n = \frac{100}{25} = 4$, i.e., a HOMO tetramer.
• SDS-PAGE interpretation: single band at a given molecular weight implies homo-oligomer with identical subunits; multiple bands imply multiple subunits or hetero-oligomeric composition.
• Diffraction basics (Fourier relation):
  • Diffraction spots provide amplitudes and phases for a Fourier series representing the electron density.
  • The observed diffraction pattern is a reciprocal-space representation of the real-space electron density ρ(r).
• Fourier synthesis (conceptual): from a set of observed spots, reconstruct the complex periodic function (electron density) by summing Fourier components:
  $\rho(\mathbf{r}) \,\approx\, \sum_{hkl} F(hkl) \, e^{2\pi i (h x + k y + l z)} ,$
  where F(hkl) encodes amplitude and phase information inferred from the data.
• R-factor (model validation): $R = \frac{\sum<em>{hkl} |F</em>{obs} - F<em>{calc}|}{\sum</em>{hkl} |F_{obs}|}$
  • Acceptable values: typically (R \lesssim 0.20) for good structures.
• Ramachandran plot (validation of backbone geometry): allowed regions for φ (phi) and ψ (psi) dihedral angles; outliers lead to rejection.
• Peptide bond planarity: due to resonance, the peptide bond has restricted rotation and is effectively planar.
• AlphaFold context (AI + data): AI-based predictions can closely match experimental structures, especially when robust data exist in structure databases; still, experimental validation remains important.

Connections to broader themes

• Foundational math in biology: Fourier analysis is not just abstract math; it directly enables deciphering atomic structures from diffraction data.
• Experiment vs computation: The lecture juxtaposes traditional experimental methods (X-ray crystallography) with modern AI approaches (AlphaFold), highlighting how science advances through both data and computation.
• Ethical/philosophical note: While AI can predict structures quickly, the integrity of science depends on validation against real data and transparent methods for peer review.

Study tips inspired by the talk

• Practice interpreting SDS-PAGE results for oligomeric states by working through subunit-mass examples and total protein mass.
• Understand the difference between what gel filtration reveals (quaternary state in native conditions) and what SDS-PAGE reveals (subunit composition under denaturing conditions).
• Review the crystallography workflow from crystallization (sitting drop) to diffraction pattern, to electron density, to model building, then refinement and validation (R-factor, Ramachandran plot).
• Relate the diffraction spots to the reciprocal lattice and unit cell dimensions; remember the analogy: diffraction is like a rainbow of spots, each spot reflecting electron density contributions from the atoms.
• Keep in mind the role of Fourier's theorem in turning a complex crystal into a sum of waves whose amplitudes and phases recreate the observed structure.
• Watch for the caveats of crystallography: crystals are solid-state snapshots that may introduce artifacts; comparison with solution-state methods (like NMR, cryo-EM) can provide complementary information.
• Stay aware of the evolving landscape with AlphaFold: use AI predictions as guidance but rely on experimental data for validation and for understanding dynamic aspects not captured in a static model.