Interpreting Biological Evidence in Court

A likelihood ratio is the answer to one precise question: given the evidence I observed in the laboratory, how much more probable is it that my observation arose under hypothesis H1 than under hypothesis H2? In a DNA case, H1 is typically 'the suspect is the source of the biological material' and H2 is 'an unknown person unrelated to the suspect is the source'. The LR is the probability of the evidence given H1 divided by the probability of the evidence given H2.

For a full single-source autosomal STR profile, the probability of the evidence given H1 is effectively 1: if the suspect is the source, we expect to see exactly their profile. The probability of the evidence given H2 is the product of the population frequencies of each allele pair across all loci, drawn from a relevant population database. A 20-locus profile might yield a random match probability of 1 in 10 billion, giving an LR of approximately 10 billion. This is the figure the expert presents to the court, not the phrase 'the suspect definitely left this sample'.

The choice of population database affects the denominator and therefore the LR. A South Asian population database will produce different allele frequencies than a West African or Northern European database. When the suspect's ancestry is known, the expert uses the most appropriate available database; when it is not, multiple databases may be run and the results reported. Indian casework increasingly draws on the NCRB population reference datasets, though these remain less comprehensive than the CODIS-based US databases or the UK National DNA Database reference sets.

Crime scene biological samples frequently contain DNA from more than one contributor, producing a mixture profile. Older interpretation methods applied binary rules: if the suspect's alleles were all present in the mixture, they were included; if one allele was absent, they were excluded. These methods struggled with complex mixtures involving three or more contributors, low-template samples where allele drop-out is possible, and samples with high stutter. They also produced non-quantified conclusions that courts could not easily weigh against each other.

Probabilistic genotyping addresses this by modelling the full probability distribution over all possible genotype combinations across all possible contributors, accounting for known artefacts such as stutter, allele drop-in (the appearance of a spurious allele), and allele drop-out (the absence of an expected allele in a low-template sample). The output is a likelihood ratio rather than a categorical inclusion. STRmix, developed in New Zealand and Australia, is validated for use in the UK, US, Canada, and Australia. TrueAllele from Cybergenetics is used across US laboratories. The ENFSI DNA Working Group has published validation and reporting guidelines covering all major platforms.

Courts have increasingly required that the probabilistic genotyping software itself be subject to disclosure and scrutiny. In the United States, several post-conviction challenges have centred on the proprietary nature of software algorithms (particularly TrueAllele) and the defendant's right to inspect the code underlying the LR that helped convict them. UK courts applying R v Dlugosz [2013] EWCA Crim 2 require that any statistical approach be adequately validated before results are admitted. The validation question is not whether the software is commercial; it is whether the specific method has been tested on samples representative of the case type.

A forensic biology report that says 'the blood matched the defendant' is incomplete and potentially misleading. Modern reporting standards require the expert to state the findings in relation to competing hypotheses, quantify the evidential weight where possible, disclose the relevant assumptions and caveats, and use language calibrated to the strength of the evidence.

The Forensic Science Regulator in England and Wales publishes Codes of Practice that are legally binding on forensic science providers following the Forensic Science Regulator Act 2021. The codes require ISO/IEC 17025 accreditation, evaluative reporting framed around defence and prosecution hypotheses, and explicit disclosure of uncertainty. In the United States, the OSAC Forensic Biology Subcommittee publishes standards documents that are not legally binding but are increasingly referenced in Daubert hearings. At the international level, the ENFSI Guideline for Evaluative Reporting in Forensic Science (2015) provides the most widely referenced framework for how LRs should be reported across member states of the European Union.

India's forensic biology reporting does not yet have a single binding national standard equivalent to the Forensic Science Regulator Codes. The Bharatiya Sakshya Adhiniyam 2023 (which replaced the Indian Evidence Act 1872) governs admissibility: expert opinion evidence is admitted when the court requires scientific knowledge beyond its own. The weight accorded to that opinion is at the court's discretion. Proposals for a national forensic accreditation body have been under development since the National Forensic Science University Act 2020, and several state laboratories are ISO/IEC 17025 accredited, but uniform evaluative reporting standards remain aspirational at national level.

Biological evidence is evidence of presence, not evidence of action. A DNA profile establishes that biological material from a given individual was found at a location. It does not establish when it was deposited, what the person was doing when they deposited it, or whether the deposition was connected to the offence under investigation. Courts regularly require the expert to address this distinction explicitly, and failure to do so has resulted in successful appeals against conviction.

Touch DNA, derived from epithelial cells shed during skin contact, presents particular challenges. Experimental studies have shown that secondary transfer of DNA (from surface A to surface B without direct contact by the profiled person) can occur through handshakes, shared objects, or fabric contact. Tertiary transfer has also been demonstrated experimentally. An LR of millions means nothing to a court if the defence hypothesis is 'the DNA was transferred innocently' rather than 'a random unrelated person is the source', because the LR as calculated does not address the transfer question at all. The expert must identify which hypothesis set the LR actually addresses.

Sample degradation reduces the number of detectable alleles and increases the likelihood of allelic drop-out. Partial profiles carry lower statistical weight, and the LR must reflect this: a six-locus partial profile from a degraded bone sample will yield a lower LR than a full 20-locus profile from fresh blood. Contamination at any stage from collection through analysis can introduce foreign alleles. The chain of custody record and the laboratory contamination exclusion process are therefore integral to the expert's conclusion and must be included in the report.

An expert witness in forensic biology occupies a specific role: to assist the court by providing scientific knowledge the court itself does not have. This role differs from that of an advocate. The expert's duty is to the court, not to the party that instructed them, a principle stated explicitly in CPR Part 35 (England and Wales), Federal Rule of Evidence 702 (US), and equivalent provisions. In India, under the Bharatiya Nagarik Suraksha Sanhita 2023 (which replaced the CrPC), courts may summon experts and direct them to produce reports; the expert's obligation of honesty to the court is embedded in the oath taken before testimony.

Effective expert testimony in biological evidence cases requires the expert to translate technical concepts without distorting them. Judges and juries do not need to understand the chemistry of STR amplification; they need to understand what the LR means, what assumptions underlie it, and what it does not address. Analogies can help, but they can also mislead. Saying 'this is like a fingerprint match' to explain a DNA LR ignores the fact that fingerprint comparison has a different statistical framework (or no formal statistical framework at all, depending on the jurisdiction).

Cross-examination of forensic biology evidence typically targets four areas: the validity of the laboratory method, the handling of the specific sample, the assumptions built into the LR calculation, and the scope of the conclusion the expert is willing to support. A well-prepared expert anticipates all four lines and answers within the honest scope of their findings. Refusing to speculate beyond the data is not a weakness in testimony; it is the expert fulfilling their role correctly.

Overstatement of biological evidence has contributed to wrongful convictions in multiple jurisdictions. Reviews by the US National Commission on Forensic Science, the UK Forensic Science Regulator, and the Australian Institute of Forensic Science have each identified patterns: experts using absolute language ('the DNA proves'), presenting random match probabilities without the denominator hypothesis, failing to disclose that a profile is a partial or mixed result, or asserting conclusions that go beyond what the analysis supports.

Prevention operates at three levels. At the laboratory level, peer review of reports before they are disclosed, with a specific check for overstatement language, catches many errors before they reach court. At the accreditation level, ISO/IEC 17025 requires documented review procedures, and auditors assess report quality. At the courtroom level, adversarial cross-examination and the availability of defence experts are the primary safeguard; this requires that defence teams have access to the underlying data, the software validation records, and the laboratory's standard operating procedures.

The correction of historical overstatement is an active legal issue. In Australia, the Victorian Institute of Forensic Medicine and the Office of Public Prosecutions have run systematic reviews of closed cases after reporting failures in hair microscopy and other biological evidence disciplines. In the UK, the Criminal Cases Review Commission has referred cases to the Court of Appeal where expert evidence was subsequently found to have exceeded what the science supported. Forensic biologists entering the field today inherit the obligation to maintain standards that prevent future corrections from being necessary.