Computer science: TestAS practice testlets

Emergency departments cannot treat patients strictly in order of arrival: a patient with a heart attack must be seen before one with a sprained ankle, even if the latter arrived earlier. Triage systems assign each patient a priority level based on a short, standardised assessment. Many of them are written as algorithms, precise sequences of decisions that always lead to the same result for the same input. Algorithms reduce variation between staff members, make decisions traceable, and can be implemented in software.

An algorithm consists of steps and decisions. In a flowchart, steps are drawn as rectangles and decisions as diamonds with two exits, yes and no. The order matters: a triage algorithm checks the most dangerous conditions first, so that a life-threatening problem is never missed because a less important question was asked earlier. Each patient follows a single path through the chart and leaves it at the first level whose condition is met.

The simplified algorithm below assigns one of five levels, from level 1 (immediate treatment) to level 5 (non-urgent). It is written in pseudocode, a structured mixture of natural language and programming constructs. The keyword IF introduces a condition, and ELSE IF is checked only if all previous conditions were false. The first condition that is true determines the result.

IF the patient is not breathing or has no pulse THEN level 1. ELSE IF the patient is unconscious, or oxygen saturation is below 90%, THEN level 2. ELSE IF pain score ≥ 8 (on a scale of 0 to 10) or heart rate > 120 per minute THEN level 3. ELSE IF temperature ≥ 39.0 °C or pain score ≥ 4 THEN level 4. ELSE level 5.

When algorithms are tested, typical inputs and borderline cases are checked. Borderline cases are values exactly at a threshold, such as a pain score of exactly 8, because errors often arise from confusing 'greater than' with 'greater than or equal to'. Testing also asks whether every possible input leads to exactly one result; an algorithm with ELSE IF chains and a final ELSE guarantees this. Finally, the order of the conditions is checked, because swapping two conditions can change the result for patients who meet both.

DNA sequences are stored and analysed as strings of the letters A, C, G and T, written by convention from the 5' end to the 3' end. Bioinformatics, the use of computers to analyse biological data, relies on a few basic operations on such strings.

Because the two strands of DNA are complementary and run in opposite directions, the sequence of the other strand is found by taking the reverse complement: each base is replaced by its partner (A with T, C with G), and the result is read in reverse order so that it too runs from 5' to 3'. The reverse complement of 5'-AACG-3' is therefore 5'-CGTT-3'. Some sequences are their own reverse complement; such palindromic sequences are typical recognition sites of restriction enzymes, which cut DNA at specific short motifs. The enzyme EcoRI, for example, cuts at the sequence GAATTC.

The GC content, the percentage of G and C bases in a sequence, differs between organisms and between regions of a genome. Because G-C pairs are held together by three hydrogen bonds and A-T pairs by two, DNA with a higher GC content needs a higher temperature to separate its strands.

To compare two sequences of equal length, the simplest measure is the Hamming distance, the number of positions at which the letters differ. Real comparisons must also allow for insertions and deletions and use alignment algorithms that insert gaps where needed. Programs such as BLAST search huge databases for sequences similar to a query, which helps to identify genes and to infer their function from related, better-studied genes.

Searching for a short pattern, or motif, in a long sequence is called string matching. The simplest method checks every possible starting position in turn. In a sequence of length n, a motif of length m can start at n − m + 1 positions. More efficient algorithms skip positions that cannot match, which matters when genomes with billions of bases are searched.

Sequence data are produced at enormous scale. A single modern sequencing machine can read billions of bases in a day, and public databases now hold sequences from hundreds of thousands of species. Before analysis, raw reads must be checked for quality and assembled into longer sequences, and every step relies on the simple string operations described here.