With the evolution of the field of forensics, greater precision, speed, and sensitivity are some factors that are taken care to process more complex evidence. For a real-life example, for a bank robbery, a perpetrator uses a pen to take down notes which is passed to the teller. In this instance, the perpetrator leaves skin cells on the pen, besides other people who did it. This results in a complex mixture of DNA from all individuals who touched the pen.
On the other hand, 30 years ago forensic scientist would probably require the bank robber to leave some drops of blood on the pen to create a DNA profile, which, today requires only a few cells.
The ability to examine biological evidence comprising low concentrations of DNA or complex mixtures is possible mainly because of computer software that use probabilistic modeling approach. These computer algorithms use complex algorithms to gauge probabilities associated with individuals that contribute to the DNA mixture.
However, in spite of novel approaches for interpretation of DNA mixture, there are limitations. This is largely due to the overall complexity of the mixture and lack of resources such as computational time, power, and cost.
To present a solution to this, researchers at the Forensic and National Security Sciences Institute have invented a new hybrid machine learning technique for mixture analysis. The method combines the virtues of current computational approach and expert analysis with those in AI and data mining.
Potentially,the new hybrid machine learning enables rapid as well as automated deconvolution of DNA mixtures with increased accuracy as compared to current methods. The advantage of the software is minimal requirement of computing and financial resources, and result into increasingly informative and high confidence conclusions.
