Recovering missing proteins based on biological complexes

Participants: Wilson Wen Bin Goh, Weijia Kong, Hui Peng, Limsoon Wong

Background

Advancements in proteomics are important to biological and clinical research because assaying protein identities/quantities paints an immediate picture of the underlying molecular landscape. However, proteomics still suffers from incomplete proteome coverage issues (i.e. not all proteins in a sample are observable in a single screen.) This gives rise to the "missing-protein problem" (MPP), which we define as difficulty in observing proteins in a proteome screen given that they are expected to be present. Due to MPP, efforts to extend proteome profiling in comparative or clinical studies, which require consistent protein/peptide detection (and accurate quantification across an extensive dynamic range), are rendered less effective.

MPP, as defined here, is different from the missing proteins defined by the Human Proteome Project (HPP). HPP missing proteins are proteins that have never been observed in the human proteome but are predicted to be present in the human proteome due to the presence of a functional gene sequence (e.g. based on genome assembly or partial transcriptome evidence); i.e. the HPP notion of missing proteins is not sample specific. Whereas MPP is focused on proteins that are present in a sample but not detected in a proteome screen on that sample. That is, MPP is about filling in "holes" in the proteomic profiling data of patients.

Current imputation methods (for filling holes in proteomic profiling data) rely on a large set of samples to estimate the correlation between the abundances of two or more proteins from the entire complement of reported proteins in these samples. They are inapplicable when there are too few samples. Moreover, some recent studies on imputation methods have shown they do not perform well on proteomic profiling data even when given a non-trivial number of samples.

Objectives

The three main goals of this project are:

We re-consider the conventional HPP guideline (aka the two-peptide rule) that a protein is only considered detected when there are at least two non-bested peptides detected and both uniquely map to that protein. We investigate whether and how this two-peptide rule can be relaxed without compromising the reliability of a proteomic screen.
We propose a novel ranking strategy for missing-protein recovery based on protein complexes. We postulate that protein complexes provide a good context for making inference of a protein's presence and its abundance. This postulate is a significant departure from conventional imputation-based approaches for filling holes in proteomics profiling data. Notably, it is applicable for predicting whether a protein is present even when there is only one sample. This is because the postulate implies on two reasonable hypotheses, which allow the likelihood of a candidate missing protein being actually present in a sample to be estimated based on the likelihood of its parent complexes being present in the sample: (i) The likelihood of a protein complex being present in a sample is proportional to the fraction of its constituent proteins that are reliably reported to be present in that sample; and (ii) the presence of a protein complex in a sample implies the presence of all its constituent proteins in that sample. The likelihood of a protein being present in a sample can then be derived from the likelihoods of the presence of protein complexes that it is a constituent of.
We propose that the abundance of a missing protein (which is predicted to be present in a sample) can be imputed from the abundance of proteins which are in the same complexes (which are likely to be present in that sample) as the missing proteins, as the correlation in the abundance of these proteins with the missing protein is more likely genuine.

Selected Publications

Wilson Wen Bin Goh, Limsoon Wong. Integrating networks and proteomics: Moving forward. Trends in Biotechnology, 34(12):951--959, December 2016. PDF
Longjian Zhou, Limsoon Wong, Wilson Wen Bin Goh. Understanding missing proteins: A functional perspective. Drug Discovery Today, 23(3):644-651, March 2018.
Weijia Kong, Bertrand Jernhan Wong, Huanhuan Gao, Tiannan Guo, Xianming Liu, Xiaoxian Du, Limsoon Wong, Wilson Wen Bin Goh. PROTREC: A probability-based approach for recovering missing proteins based on biological networks. Journal of Proteomics, 250:104392, January 2022. PDF, PROTREC Server
Zelu Huang, Weijia Kong, Bertrand Jernhan Wong, Huanhuan Gao, Tiannan Guo, Xianming Liu, Xiaoxian Du, Limsoon Wong, Wilson Wen Bin Goh. Proteomic datasets of HeLa and SiHa cell lines acquired by DDA-PASEF and diaPASEF. Data in Brief, 41:107919, April 2022. PDF
Bertrand Jernhan Wong, Weijia Kong, Limsoon Wong, Wilson Wen Bin Goh. Resolving missing protein problems using functional class scoring. Scientific Reports, 12(1):11358, July 2022. PDF
Wilson Wen Bin Goh, Weijia Kong, Limsoon Wong. Evaluating network-based missing protein prediction using p-values, Bayes factors, and probabilities. Journal of Bioinformatics and Computational Biology, 21(1):2350005, March 2023. PDF
Hui Peng, Limsoon Wong, Wilson Wen Bin Goh. ProInfer: An interpretable protein inference tool leveraging on biological networks. PLoS Computational Biology, 19(3):e1010961, March 2023. PDF, ProInfer in Scala
Wilson Wen Bin Goh, Harvard Wai Hann Hui, Limsoon Wong. How missing value imputation is confounded with batch effects and what you can do about it. Drug Discovery Today, 28(9):103661, September 2023. PDF
Weijia Kong, Bertrand Jern Han Wong, Harvard Wai Hann Hui, Kai Peng Lim, Yulan Wang, Limsoon Wong, Wilson Wen Bin Goh. ProJect: A powerful mixed-model missing value imputation method. Briefings in Bioinformatics, 24(4):bbad233, July 2023. PDF

Dissertations

Selected Presentations

Limsoon Wong. Improving coverage and consistency of MS-based proteomics. Invited talk at 12th Korea-Singapore Joint Workshop on Bioinformatics and Natural Language Processing, KAIST, Daejeon, Korea, 22-23 September 2016. PPT
Limsoon Wong. Improving coverage and consistency of MS-based proteomics. Invited keynote at 16th IEEE International Conference on Bioinformatics and Bioengineering, Taichung, Taiwan, 31 October - 2 November 2016. PPT
Limsoon Wong. Advancing clinical proteomics via analysis based on biological complexes. Keynote at 16th International Conference on Bioinformatics (InCoB), Shenzhen, China, 20 - 22 September 2017. PPT
Limsoon Wong. Robustness of protein complex-based analysis of proteomics data. Invited talk at 1st Westlake Symnposium for Proteomic Big Data, Westlake University, Hangzhou, China, 30 June 2019. PPT

Acknowledgements

This project is supported in part by MOE Tier-2 grant MOE2019-T2-1-042.

Last updated: 12/8/2023, Limsoon Wong.