Calculating Guilt: Using open-source software in forensic DNA testing
summary
Claims that probabilistic genotyping turns messy mixtures into quantitative evidence and that open source tools make that transparent. Uses Lab Retriever (validated locally) to incorporate dropout rates, raising a likelihood from 1.160 sextillion on pristine data to an intermediate 730 quintillion when dropout is modeled, compared with an understated result of about 1.4 million when weak data are discarded. Describes validation workload, limits (70 unit detection, full profiles from ~five cells), and contrasts free semi‑continuous tools with costly closed‑source continuous software like STRmix.
AI-written summary · about
chapters (15)
- 0:00Start
- 1:13Gameplan
- 2:14Rosalind was robbed.
- 3:43STR = Short Tandem Repeat
- 5:59You are not a special snowflake
- 15:40Included or excluded?
- 19:40Likelihood Ratio (LHR)
- 22:51The world is a dirty place.
- 24:00A wretched hive of scum and villainy
- 24:50A reasonable degree of scientific certainty.
- 28:29The loci with 2 minor alleles
- 30:02Probabilistic genotyping Semi-continuous
- 31:21Lab Retriever
- 34:34Probabilistic genotyping Continuous
- 40:25Statistics are hard.
related talks
29mAnalysing and sharing genetic data with Python
30mAlgorithms for DNA Data Storage
26mBuilding a scalable genomic matching algorithm
39mChallenges of analysing the wheat genome
57mCode Genome - Fingerprinting Code to Build Trustworthy SBOMs
17mOpen Source for Data Quality
41mWhy would you store digital information in DNA?
11mTracking Genome Sequence Data of Cancer Patients with Docker