티스토리 뷰

Isobaric labeling proteomics에서 “정확한 정량”이라는 믿음이 흔들리는 순간
Proteomics에서 iTRAQ와 TMT가 등장했을 때 많은 연구자들은 거의 혁명처럼 받아들였다. 서로 다른 sample을 하나의 mixture로 합친 뒤 동시에 LC-MS/MS 분석을 수행할 수 있고, label-free quantification에서 반복적으로 문제가 되었던 run-to-run variability도 크게 줄일 수 있었기 때문이다.
특히 multiplexing capability는 enormous advantage였다. 6-plex, 10-plex, 16-plex를 넘어 최근에는 더 높은 plexing 전략까지 가능해지면서 clinical cohort와 large-scale comparative proteomics 연구가 훨씬 효율적으로 진행되기 시작했다.
무엇보다 많은 사람들이 가장 매력적으로 느낀 부분은 “정량 정확도”였다. Sample을 각각 따로 주입하는 LFQ와 달리, iTRAQ/TMT는 서로 다른 sample peptide를 같은 LC-MS run 안에서 동시에 측정한다. 따라서 chromatographic fluctuation이나 spray instability 영향이 줄어들고 더 정확한 quantification이 가능할 것처럼 보였다.
실제로 abundance ratio는 매우 깔끔하게 계산된다. Reporter ion intensity를 비교하면 sample 간 relative abundance가 숫자로 정리된다. Volcano plot과 heatmap은 더욱 선명해지고, pathway difference는 마치 biology가 직접 드러난 것처럼 보인다.
하지만 raw data를 오래 보다 보면 어느 순간 이상한 현상이 반복된다. 실제 biological difference보다 fold change가 훨씬 압축되어 보이고, low abundance signaling protein은 기대보다 변화가 작게 나타나며, 어떤 peptide는 분명 differential signal이 있어야 하는데 거의 ratio 1 근처로 수렴한다.
처음에는 normalization 문제처럼 느껴질 수 있다. Search parameter 설정 때문처럼 보일 수도 있다. 그러나 데이터를 끝까지 따라가다 보면 결국 하나의 핵심 구조에 도달하게 된다.
iTRAQ/TMT는 “같이 측정한다”는 장점 때문에 오히려 새로운 종류의 quantification distortion을 만들어낸다는 점이다.
그리고 이 구조를 이해하기 시작하면 이전에는 너무 당연하게 믿었던 reporter ion ratio가 전혀 다르게 보이기 시작한다.
1. iTRAQ/TMT는 peptide를 섞어서 측정한다
iTRAQ와 TMT의 기본 개념은 매우 elegant하다.
각 sample peptide에 서로 다른 isobaric tag를 붙인 뒤 모든 sample을 하나로 섞는다. MS1 단계에서는 동일 mass로 보이기 때문에 co-eluting single precursor처럼 검출된다. 이후 fragmentation이 일어나면 reporter ion이 분리되어 sample-specific intensity를 제공한다.
즉 quantification은 MS2 reporter ion intensity 기반으로 이루어진다.
문제는 바로 여기서 시작된다.
Mass spectrometer는 특정 precursor isolation window 안에 들어온 ion을 함께 fragmentation한다. 그런데 실제 biological sample에서는 하나의 precursor만 완벽하게 isolation되는 경우가 드물다.
결국 target peptide 외의 다른 co-isolated peptide들도 함께 fragmentation에 들어간다.
그리고 이 contamination이 reporter ion ratio 자체를 바꾸기 시작한다.
2. Co-isolation interference가 ratio를 압축한다
이 현상은 iTRAQ/TMT 정량 왜곡의 핵심이다. 흔히 ratio compression이라고 부른다.
예를 들어 Sample A에서 어떤 peptide abundance가 매우 높고 Sample B에서는 거의 없다고 가정해보자. 이상적으로라면 reporter ion ratio는 극단적인 fold change를 보여야 한다.
하지만 실제 isolation window 안에는 unrelated background peptide들이 함께 들어온다. 이 contaminating peptide들은 여러 sample에서 평균적인 abundance를 가지는 경우가 많다.
결과적으로 reporter ion signal에 background contribution이 추가된다.
즉 원래 10:1이어야 할 ratio가 3:1 혹은 2:1처럼 압축되어 보이기 시작한다.
실제 TMT dataset에서는 strong biological regulation조차 surprisingly modest fold change로 나타나는 경우가 많다.
그리고 문제는 이 압축이 random하지 않다는 점이다. Low abundance peptide일수록 interference 영향을 훨씬 크게 받는다.
3. Low abundance protein은 interference에 가장 취약하다
High abundance peptide는 target signal 자체가 강하기 때문에 일부 contamination이 들어와도 상대적으로 영향이 적다.
하지만 low abundance peptide는 상황이 다르다.
Target precursor intensity가 약한 상태에서 co-isolated background ion이 조금만 섞여도 reporter ion composition이 급격히 왜곡된다.
특히 signaling protein, transcription factor, phosphoprotein 같은 low abundance biology는 원래 signal 자체가 작기 때문에 ratio compression 영향을 극단적으로 받는다.
실제 phosphoproteomics TMT dataset에서는 중요한 phosphosite regulation이 거의 fold change 1 근처로 flattening되는 경우가 자주 관찰된다.
즉 biologically 중요한 subtle regulation이 technical interference에 의해 희석된다.
4. Isolation window는 완벽하게 깨끗하지 않다
많은 사람들이 quadrupole isolation이 매우 precise하다고 생각한다. 하지만 실제 LC-MS 환경은 생각보다 훨씬 crowded하다.
특히 complex proteome sample에서는 동일 retention time에 수많은 peptide precursor가 함께 존재한다.
예를 들어 0.7~1.0 m/z isolation window 안에도 여러 precursor tailing signal이 동시에 들어올 수 있다. Co-eluting peptide density가 높을수록 interference probability는 급격히 증가한다.
실제 Orbitrap raw data를 보면 isolation purity가 매우 낮은 precursor도 흔하게 발견된다.
문제는 software abundance table에서는 이런 contamination 구조가 거의 보이지 않는다는 점이다.
연구자는 reporter ion ratio만 보게 되고, 그 ratio가 얼마나 혼합된 signal인지 알기 어려워진다.
5. MS2 기반 quantification 자체의 구조적 한계
iTRAQ/TMT quantification은 기본적으로 MS2 fragmentation에 의존한다.
문제는 fragmentation efficiency 자체가 precursor composition 영향을 받는다는 점이다.
Co-isolated peptide가 많아지면 fragment ion complexity가 증가하고 reporter ion intensity distribution도 흔들린다.
또 precursor intensity가 매우 낮으면 reporter ion signal 자체가 noisy해진다.
특히 low abundance peptide는 reporter ion signal-to-noise ratio가 급격히 나빠질 수 있다.
즉 iTRAQ/TMT abundance는 단순 protein amount가 아니라 fragmentation environment 전체의 영향을 함께 받는다.
6. Ratio compression은 biology를 flattening한다
이 문제가 가장 위험한 이유는 실제 biological difference를 systematically 줄여버린다는 점이다.
예를 들어 disease sample에서 특정 signaling pathway가 강하게 활성화되었다고 가정해보자. 실제 biological fold change는 매우 클 수 있다.
하지만 TMT interference가 reporter ion ratio를 압축하면 downstream analysis에서는 moderate regulation 정도로 보일 수 있다.
결과적으로 biologically 중요한 pathway activation이 underestimated된다.
실제 comparative study에서는 western blot이나 targeted MS에서는 strong change가 보이는데, TMT global proteomics에서는 surprisingly weak fold change만 나타나는 사례가 자주 보고된다.
즉 iTRAQ/TMT는 differential signal을 “부드럽게 평탄화”하는 경향을 가진다.
7. SPS-MS3도 완벽한 해결책은 아니다
이 문제를 해결하기 위해 등장한 것이 SPS-MS3 전략이다.
MS2 fragment ion 일부를 다시 isolation해서 MS3 reporter ion quantification을 수행하면 co-isolation interference를 줄일 수 있다.
실제로 SPS-MS3는 ratio compression 완화에 상당한 효과를 보인다.
하지만 완전히 해결되지는 않는다.
MS3 acquisition은 scan cycle time을 증가시키고 sensitivity를 희생할 수 있다. 특히 low abundance peptide는 MS3 단계까지 충분한 signal을 유지하지 못하는 경우가 있다.
또 precursor selection 자체가 imperfect하면 residual interference가 남는다.
즉 SPS-MS3는 distortion을 줄이지만, iTRAQ/TMT quantification을 완전히 “순수한 abundance measurement”로 만들지는 못한다.
8. Batch effect는 여전히 존재한다
많은 사람들이 TMT multiplexing이 batch effect를 제거한다고 생각한다.
하지만 실제로는 batch effect가 사라지는 것이 아니라 형태가 바뀐다.
한 TMT set 안에서는 relative quantification consistency가 높을 수 있다. 하지만 multiple TMT batch를 연결하기 시작하면 bridge sample normalization과 batch alignment 문제가 다시 등장한다.
또 labeling efficiency variability, mixing error, sample loading difference 같은 요소도 batch dependency를 만든다.
실제 large-scale TMT cohort study에서는 TMT batch structure가 PCA clustering을 지배하는 경우도 적지 않다.
즉 multiplexing은 variability를 줄여주지만 biology와 technical structure를 완전히 분리해주지는 않는다.
9. Reporter ion intensity는 “진실”처럼 보인다
iTRAQ/TMT가 특히 위험한 이유는 결과가 매우 깔끔하게 보인다는 점이다.
Reporter ion table은 숫자가 정돈되어 있고 missing value도 LFQ보다 적다. Heatmap은 선명하고 clustering도 아름답게 나온다.
그래서 연구자는 quantification이 매우 accurate하다고 느끼기 쉽다.
하지만 실제로는 reporter ion intensity 뒤에 co-isolation interference, ratio compression, precursor purity, fragmentation variability 같은 복잡한 요소들이 숨어 있다.
그리고 대부분의 distortion은 abundance table 안에서 invisible하다.
즉 우리는 “정확한 정량”이라고 믿고 있지만, 실제로는 compressed biological reality를 보고 있을 수도 있다.
10. 실제 데이터에서 반드시 확인해야 하는 것들
TMT/iTRAQ 데이터를 해석할 때는 isolation interference level을 반드시 확인할 필요가 있다.
가능하다면 precursor isolation purity metric을 함께 보는 것이 좋다. Low purity precursor는 ratio distortion 가능성이 높다.
Low abundance protein과 phosphopeptide는 특히 cautious interpretation이 필요하다. Small fold change가 실제 biology에서는 훨씬 큰 regulation일 가능성도 있다.
SPS-MS3 사용 여부와 acquisition setting 역시 중요하다. Orbitrap resolution, AGC target, injection time은 reporter ion quality에 직접 영향을 준다.
또 orthogonal validation이 매우 중요하다. Western blot, PRM/SRM, DIA verification 같은 independent approach를 통해 key finding을 재확인할 필요가 있다.
특히 signaling pathway 해석에서는 “작은 fold change = 약한 biology”라고 단순하게 결론내리지 않는 것이 중요하다.
결론
iTRAQ/TMT는 proteomics 역사에서 매우 강력한 기술이었다. Multiplexing과 improved reproducibility는 large-scale comparative proteomics를 크게 발전시켰다.
하지만 isobaric labeling quantification은 단순한 “정확한 abundance measurement”가 아니다. 실제로는 co-isolation interference, ratio compression, precursor purity, fragmentation complexity 같은 요소들이 함께 작동하는 composite measurement에 가깝다.
즉 reporter ion ratio는 biology 자체를 보여주는 동시에 LC-MS crowded environment 안에서 살아남은 signal 구조까지 함께 반영한다.
이 사실을 이해하기 시작하면 이전에는 너무 깔끔하게 보였던 TMT abundance table이 다르게 보이기 시작한다. 왜 strong biology가 weak fold change처럼 보이는지, 왜 low abundance signaling pathway가 flattening되는지, 왜 orthogonal validation에서 더 큰 차이가 나타나는지에 대한 답이 isobaric quantification 구조 안에 숨어 있는 경우가 생각보다 많기 때문이다.
'제약산업' 카테고리의 다른 글
| Label-free quantification의 한계와 착각 (0) | 2026.05.23 |
|---|---|
| Freeze-thaw가 proteome에 미치는 영향 (0) | 2026.05.22 |
| Batch 간 sample prep 차이가 만드는 오류 (0) | 2026.05.21 |
| Protein precipitation이 특정 단백질을 잃게 만드는 이유 (0) | 2026.05.20 |
| Low abundance protein이 사라지는 이유 (0) | 2026.05.19 |
| Sample cleanup이 bias를 만드는 이유 (0) | 2026.05.18 |
| Trypsin digestion이 완벽하지 않은 이유 (0) | 2026.05.17 |
| Digestion 효율이 quantification에 미치는 영향 (0) | 2026.05.16 |
- Total
- Today
- Yesterday
- 미래산업
- Spatial metabolomics
- 임상시험
- metabolomics
- LC-MS
- Targeted Metabolomics
- 데이터
- Biomarker
- 시스템
- 분석
- 약물분석
- biological signal
- audit
- matrix effect
- 대사체 분석
- Multi-omics
- bioanalysis
- AI
- 제약
- 분석팀
- 정량분석
- 해석
- 제약산업
- 신약개발
- 정밀의료
- 치료제
- Missing Value
- lc-ms/ms
- 바이오마커
- Proteomics
| 일 | 월 | 화 | 수 | 목 | 금 | 토 |
|---|---|---|---|---|---|---|
| 1 | 2 | |||||
| 3 | 4 | 5 | 6 | 7 | 8 | 9 |
| 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| 17 | 18 | 19 | 20 | 21 | 22 | 23 |
| 24 | 25 | 26 | 27 | 28 | 29 | 30 |
| 31 |
