Thanks Nicole! ☺️

😷 We also investigated the source of resistance in samples with resistant subpopulations using within-sample diversity, which suggests secondary infection events in at least 28% of samples.

⌛ Building on my previous work on compensatory mutations, we found that samples with resistant subpopulations were less likely to be compensated than homogeneous resistant samples, hinting at recent resistance acquisition.

📈 Relaxing the read support threshold, for example, can increase the sensitivity of rifampicin resistance prediction to over 95%.

Key take-aways:
🚧 In this work, we show that conservative parameters in bioinformatic pipelines can obscure resistant subpopulations in clinical samples of M. tb.

Having often grappled with the limited translational aspect of my pre-PhD work, the real-world impact of the topics discussed in this session and at this conference was refreshing and motivating and will help me push my PhD at @modmedmicro.bsky.social over the finish line later this year!

🚧 We showed that this issue can be easily mitigated by lowering the read support threshold required to call variants in whole-genome sequencing pipelines. This can lead to significantly increased sensitivity for rifampicin resistance prediction.

🧬 The talk allowed me to raise a central issue in sequencing-based diagnostics for M. tuberculosis: they often fail to detect resistant subpopulations - something that can have serious clinical implications.

😷In a third of our clinical samples with resistant subpopulations, resistance was likely caused by a secondary infection instead of within-host evolution. Hence the patient contracted an M. tuberculosis infection at least twice, independently.

📈We managed to increase the sensitivity of rifampicin resistance prediction to over 95%, with no significant trade-off for specificity.