Systematic Literature Review (SLR)

• Systematic Literature Review (SLR) is a protocol-driven method that rigorously identifies and synthesizes high-quality research to support evidence-based decisions.
• It employs structured steps such as protocol definition, comprehensive database querying, and multi-stage screening to minimize bias.
• SLRs are widely applied across fields—from medicine to bioinformatics—ensuring transparency, reproducibility, and actionable insights.

A Systematic Literature Review (SLR) is a formal, protocol-driven methodology for identifying, evaluating, and synthesizing all available, high-quality research relevant to a specific question or set of objectives. Distinguished by stringent inclusion/exclusion criteria, multi-stage screening, and transparent documentation, the SLR framework aims to minimize bias, maximize reproducibility, and support evidence-based decision-making across domains. While SLRs were first institutionalized in medicine and environmental science, their adaptation to data-driven fields such as bioinformatics has required domain-specific workflows, role structures, and iterative quality controls (Mariano et al., 2017).

1. Process Structure and Methodological Principles

The SLR process is commonly organized into four iterative, interdependent phases, as formalized in the BiSLR (Bioinformatics Systematic Literature Review) framework:

1. Protocol Definition (Planning): Establish the review's main question, objectives, detailed eligibility rules, and reviewer roles before any literature search is initiated.
2. Reference Collection (Identification): Systematically query relevant bibliographic databases using precisely formulated search strings.
3. Data Evaluation (Study Selection and Quality Assessment): Progressively narrow the candidate set of references via structured, multi-level screening and scoring.
4. Interpreting (Synthesis and Reporting): Extract critical data from included studies, synthesize results (quantitatively or narratively), and present findings via tables, figures, and written analysis (Mariano et al., 2017).

Each phase operates in a spiral: if, at any stage, predefined objectives are unmet, the process revisits and refines earlier steps.

2. Protocol Design, Criteria, and Documentation

Protocol definition is foundational. The SLR protocol must specify:

• A singular, focused main question (e.g., “Which bioinformatics tools support large-scale gene array analysis in heart failure?”).
• 1–5 “specific questions” that further break down the main theme (e.g., time window, data types, validation approaches).
• Inclusion criteria (e.g., must mention systematic review methodology, cover domain-specific themes, or describe relevant databases/tools).
• Exclusion criteria (e.g., non-English publications, unverifiable results, absence of validation, sub-threshold quality scores).
• Team structure, with ≥2 (and ideally ≥4) independent reviewers to reduce bias; clearly assigned drafting, searching, screening, and conflict-resolution roles.

Documentation is continuous: the protocol should be stored in a structured table (specifically Table 1 in (Mariano et al., 2017)), and optionally registered in a public registry (e.g., PROSPERO) to ensure process transparency and reproducibility.

3. Systematic Identification and Data Handling

The reference collection phase demands:

• Querying ≥2 domain-relevant databases (e.g., PubMed, ScienceDirect, Scopus, Web of Science) using search strings derived directly from the protocol's main and specific questions.
• Use of exact-phrase, Boolean operators, and field tags within search queries; iterative refinement and yield-tracking are mandatory (see Supplementary Table S2 of (Mariano et al., 2017)).
• Export of all candidate results and aggregation via reference-management tools or scripts.
• De-duplication by unique identifiers (preferably DOI) using automated scripts.
• Verification of coverage against the protocol’s objectives, with considered revision of queries if gaps remain.

Systematic recording of search iterations and results is enforced to support auditability, as is the generation of a PRISMA-style flow diagram to track each candidate’s journey from retrieval to inclusion.

4. Multi-Stage Screening, Quality Assurance, and Inclusion Rules

Data evaluation proceeds as a four-step cascade:

1. Title Screening: Papers are advanced if any title keyword matches the review question; ambiguous cases default to inclusion.
2. Abstract Screening: Abstracts are excluded only if explicit violations of criteria are evident; ambiguous cases move forward.
3. Diagonal Reading: Reviewers read introductions, figures/tables, conclusions—evaluating question-addressing adequacy and possible refinement needs.
4. Full-Text Scoring: Each study is assessed per specific question, with a 0 (not met), 1 (partially met), or 2 (fully met) scale. The sum score $S = \sum_{i=1}^5 s_i$ is computed, and an inclusion threshold enforced: $S \geq 6$ (≥60%).

All screening is performed independently. Advancement at each step requires approval by ≥50% of reviewers, with group consensus used to resolve residual uncertainty.

Comprehensive logs of inclusion/exclusion—including explicit reasons for each decision—underpin the transparency and repeatability of the process. Use of structured reporting diagrams (PRISMA-style) is mandated to visualize study selection flow.

5. Data Extraction, Synthesis, and Reporting Standards

For each included study, a standardized Data Extraction Form is completed, comprising:

• Bibliographic metadata (authors, year, journal, DOI)
• Experimental design and aims
• Field/subdomain (e.g., genomics, proteomics)
• Databases and tools mentioned
• Key quantitative findings (e.g., statistical frequencies, performance metrics)
• Structured answers/scores to the protocol-specific questions

Pilot testing of extraction fields is conducted before full deployment to minimize ambiguity. The resulting datasets allow for grouping (by subdomain, methodology, or other) and thematic summarization—encompassing common patterns, gaps, or contradictory evidence. Where data support, meta-analysis or quantitative synthesis (e.g., frequency counts, network diagrams) may be performed.

Reporting is executed in tabular, graphical, and narrative formats. Figures may include PRISMA flowcharts, bar/heatmaps of domain distribution, or network visualizations of tool–database associations (e.g., via Cytoscape).

Quality assessment of the SLR as a whole employs the PRISMA checklist (documented in full as Supplementary Table S4), including critical commentary on protocol and reporting limitations at both study and SLR levels.

6. Best Practices, Automation, and Methodological Rigor

The BiSLR approach mandates continuous, protocol-driven refinement: query formulation, inclusion rules, and extraction fields are iteratively optimized based on intermediate results and group discussion. The entire process enforces dual/multi-review at each stage, with conservative inclusion bias to safeguard against premature filtering.

Structured team-based workflows, meticulous logging, and conflict resolution through independent review and group meetings are crucial safeguards against individual bias or oversight.

Deliverables include the registered protocol, the full SLR manuscript (with Methods, Results, Discussion, PRISMA diagram, tables, and figures), and—where feasible—curated domain-specific datasets for reuse.

By adhering to these structured, iterative steps, a team achieves a transparent, reproducible, and low-bias Systematic Literature Review, establishing a high standard for evidence-based practice in any technical field (Mariano et al., 2017).