BMS5021 · Introduction to Bioinformatics
Quantitative Proteomics
Week 10 of Monash University BMS5021 asks not just which proteins are present but HOW MUCH: relative versus absolute quantification, label-free versus labelled (isobaric) methods, normalisation, and testing for differential protein abundance between conditions. It also confronts the practical headache of missing values, using the SARS-CoV-2 infection timecourse as the worked case. This is the analytical method behind the Topic 3 report, so the quantification choices and missing-value handling are exactly what you will justify to markers.
What this chapter covers
- 01Aim of quantitative proteomics: relative (fold-change between conditions) or absolute (copies per cell) protein measurement
- 02Proteome fractionation reduces complexity before quantification
- 03Label-free quantification vs labelled (isobaric) quantification approaches
- 04Normalisation and testing for differential abundance between conditions (the Topic 3 analysis)
- 05Week-10 case: SARS-CoV-2 host + viral proteome/translatome timecourse of virus infection (MSFragger search output)
- 06Missing values are common in proteomics; filtering rule: keep proteins present in >=25% of samples (<=75% missing)
- 07Perseus-method imputation: replace missing values with draws from a normal distribution down-shifted by 1.8 SD, width 0.3 of each sample
- 08Why a lenient threshold matters for a timecourse: late-appearing viral proteins must not be filtered out
Justify a missing-value threshold for a viral timecourse (Week 10-12, Topic 3 reasoning)
- +1State the filter. The taught rule is to keep proteins (rows) that are present in at least 25% of samples - equivalently, with no more than 75% missing values - as a starting point (a suggested alternative is to require presence in at least 50% of samples). This removes proteins seen too rarely to analyse reliably while keeping most usable data.
- +1Justify the lenient threshold for a timecourse. In a virus-infection timecourse, viral proteins only appear at LATER timepoints, so they are legitimately absent (missing) from early samples. A strict 'present in most samples' filter would wrongly discard these late-appearing viral proteins. A lenient threshold (present in only >=25% of samples) retains them, preserving exactly the biology you are trying to study.
- +1Outline the imputation. After filtering, remaining missing values are imputed rather than left blank: the Perseus method replaces each missing value with a random draw from a normal distribution that is DOWN-SHIFTED by 1.8 standard deviations and given a WIDTH of 0.3 of that sample's distribution. This models missing values as low-abundance (near the detection limit) proteins rather than assuming they are zero or average.
Key terms
- Relative vs absolute quantification
- Relative quantification measures fold-changes in a protein's abundance between conditions; absolute quantification measures actual amounts (e.g. copies per cell or stoichiometry). Topic 3 focuses on relative, differential abundance.
- Label-free vs labelled quantification
- Label-free methods infer abundance from signal intensity or spectral counts across separate runs; labelled (isobaric) methods tag samples with mass tags so they can be mixed and compared in one run. Each has trade-offs in cost, throughput and accuracy.
- Proteome fractionation
- Separating the proteome into fractions before analysis to reduce its complexity, so lower-abundance proteins can be detected and quantified.
- Missing values (proteomics)
- Proteins undetected in some samples, common in MS data. Handled by filtering (keep proteins present in >=25% of samples) then imputation, rather than treating absence as a true zero.
- Perseus imputation
- A method that replaces missing values with random draws from a normal distribution down-shifted by 1.8 SD and with width 0.3 of each sample, modelling missing values as low-abundance proteins near the detection limit.
- Differential abundance
- The proteomics analogue of differential expression: testing which proteins change significantly in abundance between conditions after normalisation - the core output of the Topic 3 analysis.
Quantitative Proteomics FAQ
How does Week 10 relate to the Topic 3 report?
Directly - the Topic 3 report (35%) is a quantitative-proteomics / mass-spectrometry data analysis, and Week 10 supplies the quantification and differential-abundance methods, plus the missing-value handling, that the report is built on. The Week-10 SARS-CoV-2 timecourse is the kind of dataset and analysis you are assessed on. Confirm the exact report task and dataset on Moodle.
What is the difference between a zero and a missing value in proteomics?
A zero or blank in an MS dataset usually means the protein was not DETECTED in that sample, which is not the same as it being truly absent - it may simply be below the detection limit. That is why you don't treat missing values as real zeros: you filter proteins seen too rarely, then impute the rest as likely low-abundance signals (Perseus method) rather than assuming they are zero or average.
When should I use a strict versus a lenient missing-value threshold?
Match it to the experimental design. For a timecourse where proteins legitimately appear only at some timepoints (like viral proteins late in infection), use a lenient threshold (present in >=25% of samples) so you don't discard real biology. For a stable comparison where every protein should be seen in most samples, a stricter threshold (e.g. >=50% present) removes noise. Always justify the choice by the biology, and confirm expectations on Moodle.
Can AI help me with the quantitative-proteomics analysis?
Yes. Sia can explain label-free versus labelled quantification, why proteomics has so many missing values, and how the filtering and Perseus imputation steps work - and it can pressure-test the justification you plan to write for your threshold and normalisation choices. It explains and checks your reasoning; it does not run your graded Topic 3 analysis or write the report, and Monash University academic-integrity rules apply.
Exam move
Because the Topic 3 report turns on defensible analysis choices, practise justifying them, not just performing them. Be able to compare relative versus absolute and label-free versus labelled quantification and say when each fits. Master the missing-value story end to end: filter to proteins present in at least 25% of samples, justify a lenient threshold for a timecourse where viral proteins appear late, then impute remaining values by the Perseus method (down-shift 1.8 SD, width 0.3). Rehearse on the SARS-CoV-2 timecourse logic so you can narrate every step. Ask Sia to challenge your justifications and to check your reasoning against the experimental design. Confirm the report's dataset, methods and rubric on Moodle.
Working through Quantitative Proteomics in BMS5021? Sia is AskSia’s AI Biology tutor — ask any BMS5021 Quantitative Proteomics question and get a clear, step-by-step explanation grounded in how BMS5021 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
