Thread by @sonjantang

🎉Immensely grateful and overwhelmingly joyful that my 1st first-author paper is out now in @BMCMedicine😁!!!🎉

This project started as my master’s thesis in @ImperialSPH with @VerenaZuber and @KTsilidis in the summer of 2020 😱!

Tweetorial 🧵⏬!

1/11

bmcmedicine.biomedcentral.com/articles/10.11…

🎯Aim: We wanted to examine and rank all 34 blood and urine biochemical biomarkers in the @uk_biobank for possible #causal relationships with #breastcancer liability in women.

2/11

🧮Methods: Since experimenting on biomarkers in human participants poses ethical and practical questions, we instead performed a method called Mendelian Randomisation (MR)🧬.

MR uses observational data👀 and an instrumental variable (IV) framework to infer causality.

3/11

🧮Methods: MR parallels a randomised controlled trial by leveraging the random inheritance of alleles at conception to investigate the causal relationship between the genetically predicted exposure and outcome.

4/11

🧮Methods: We used a variety of methods to identify and rank our biomarkers for causal relationships with breast cancer.
💻 Main = IVW MR
💻 Sensitivity = MR-Egger, Weighted median, MR-PRESSO
💻 Additional = multivariable and bidirectional MR
🥇 Ranking = MR-BMA

5/11

📊Data: We used publicly available summary-level genome-wide association study (GWAS) data on all 34 @uk_biobank biomarkers from the @bmneale lab and GWAS data on breast cancer from the Breast Cancer Association Consortium.

6/11

📈Results: We found likely causal associations between each of the following biomarkers and higher⬆️ or lower⬇️ overall breast cancer liability:
1⃣ Testosterone = ⬆️liability
2⃣ HDL cholesterol = ⬆️liability
3⃣ IGF-1 = ⬆️liability
4️⃣ Alkaline phosphatase (ALP) = ⬇️liability

7/11

💭Discussion: We corroborate existing findings, but the ALP finding is novel! We hypothesise that our ALP-associated gene variants are associated with an enzyme that breaks down oestrogen to the less potent oestrone, thus showing a reduced risk of breast cancer.

8/11

💭Discussion: One limitation is that we limited our sample to white-British women to avoid population stratification and cannot generalise to women of other ethnicities. Also, there is still the possibility of residual pleiotropy.

9/11

💭Discussion: Some strengths include our large sample of ~420,000 women! We avoid sex-specific effects, performed many methods to scrutinise our results, and included information on the relative importance of each biomarker.

10/11

🎇Finally, our work is licensed under the @creativecommons “Attribution license”, which allows anyone to read, copy, distribute and make derivative works, as long as the authors of the original work are cited – so feel free to share, use, and cite our work widely😊!

11/11