= Solution
\b[A defensible answer is a measured lower bound for a specified biological sample, accompanied by detection limits and unresolved ambiguities.] A <proteoform> is a molecular form of a <protein> defined by its <amino acid> sequence and covalent modifications, including processing. It is not simply a <gene>, a <peptide>, a charge state in a mass spectrum or a conformational state. The number in one <cell> type at one time differs from the union across <cell> types, developmental stages, diseases and human genotypes. Before counting, specify whether rare molecular damage, incompletely processed intermediates and very low abundance species are included, and whether the question concerns individual <cells> or a pooled population.
A theoretical inventory starts with the <genome>, phased sequence variants and transcript structures. <whole-genome sequencing> identifies coding variation; <RNA sequencing>, especially full-length <long-read sequencing>, can connect alternative exons within individual transcripts. <ribosome profiling> supplies evidence that candidate <reading frames> are translated. However, an <RNA> isoform is not proof of a stable <protein>, and none of these methods alone establishes the combination of <post-translational modifications> on the mature chain. A schematic upper bound is
$$
N_{\mathrm{possible}}=\sum_i\prod_{j=1}^{m_i}(1+r_{ij}),
$$
where $i$ indexes distinct sequence forms and site $j$ permits an unmodified state plus $r_{ij}$ mutually exclusive modified states. This assumes independence between sites and that every combination is realizable. Steric constraints, processing, mutually exclusive pathways and low occupancy invalidate that assumption in real <cells>. Multiplying all documented sites therefore estimates a possible space, not the number that exists.
<bottom-up proteomics> provides a broad experimental survey. Extract <proteins> under conditions that minimize <proteolysis> and preserve labile modifications; fractionate, digest with complementary <proteases>, and analyse <peptides> by <liquid chromatography-mass spectrometry> and <tandem mass spectrometry>. Search against a sample-specific sequence database and relevant modification hypotheses, then validate sites with diagnostic fragments and, where needed, synthetic standards. Modification-specific enrichment increases sensitivity but changes the sampled population and cannot supply an unbiased count by itself. Control peptide- and protein-level <false discovery rates>, distinguish unique from shared <peptides>, and report ambiguously localized modifications explicitly. Enlarging a search to every possible modification greatly increases the chance of a convincing but incorrect match.
The central limitation is loss of linkage. Suppose sites A and B occur on different digestion products. A mixture containing equal amounts of unmodified and doubly modified chains gives 50% modification at each site. A different mixture containing equal amounts of A-only and B-only chains gives exactly the same site occupancies, but contains different intact forms. No amount of more precise measurement of these two isolated <peptides> distinguishes the populations. A spanning <peptide>, a larger linked fragment or intact-chain evidence is required.
\Image[/past-exam-of-the-biochemistry-course-of-the-university-of-cambridge/2024/paper-1-proteoform-linkage.png]
{title=Different intact proteoform populations can give identical modification frequencies after digestion}
<top-down proteomics> addresses this by separating and measuring intact <proteins> before fragmenting their selected ions. High-resolution intact mass constrains the total composition, while sequence fragments connect modifications to the same molecule and localize them. Electron-based fragmentation can help retain modifications lost during energetic fragmentation. <middle-down proteomics> uses larger fragments to recover partial linkage when complete intact-chain characterization is difficult. <John C. Tran> and colleagues demonstrated discovery-scale intact-protein analysis, illustrating that sequence variants and modification patterns can be mapped together rather than reconstructed solely from disconnected <peptides>.<Mapping intact protein isoforms in discovery mode using top-down proteomics by John C. Tran and others>[Intact-protein fractionation and identification of linked isoforms]{ref} Intact mass alone remains insufficient: positional <isomers> with the same mass require informative fragmentation or additional separation, and one envelope can contain unresolved forms.
A practical programme would combine these measurements across well-defined human <cell> populations, using biological replicates, standardized handling and a documented range of physiological perturbations. It would use sequence-resolved <RNA> data to constrain searches, broad bottom-up data to locate promising variation, and targeted top-down or middle-down analysis to establish co-occurrence. Spike-in <proteins> spanning abundances and sizes would measure extraction loss and detection sensitivity. Time courses, immediate <protein denaturation>, <protease> and <phosphatase> inhibition, and processing blanks would help distinguish cellular modification from <oxidation>, <protein deamidation> or cleavage during preparation. Very large, hydrophobic and scarce <proteins> require specific extraction and separation strategies; their absence from a spectrum cannot be treated as biological absence.
Bulk and single-cell measurements answer different questions. A bulk sample may show two <proteoforms> expressed in different <cells>; it does not show that both occur in each <cell>. Conversely, detecting two modifications in one <cell> does not demonstrate that they occupy the same <protein> molecule. Single-cell sensitivity and molecular linkage must therefore be evaluated separately. Quantitative standards and targeted follow-up can establish whether a new low-abundance assignment is reproducible rather than a one-off incorrect match.
Finally, merge identifications by molecular identity rather than by <peptide> count, search-engine entry or ion charge. Record fully characterized forms separately from groups sharing an unresolved mass or site pattern, and control identification error at the <proteoform> level. Plot discovery against sampling depth, tissues and perturbations; any extrapolation to unseen forms requires explicit assumptions about abundance and sampling heterogeneity. \b[Sequence prediction supplies candidates, <peptide> analysis supplies coverage, and intact or linked-fragment evidence establishes combinations. A universal exact count cannot be inferred by multiplying modification sites or summing database entries.]
Back to article page
