INTERPRETABLE AI FOR CHEMOTHERAPY RESPONSE
IN BREAST CANCER: AN AFRICAN PERSPECTIVE

Dr. Shamim Nabuuma Kaliisa

Breast cancer has emerged as the most frequently diagnosed malignancy among women across the African continent, presenting a clinical burden that is both substantial and growing. Current estimates indicate that Africa records over 190,000 new cases annually, characterized by mortality rates that remain disproportionately high relative to incidence (Sung et al, 2021). Within sub-Saharan Africa, the mortality to incidence ratio often exceeds 0.5, signifying that more than half of diagnosed patients succumb to the disease, a stark disparity compared to high income regions (Bay et al, 2018).

This crisis is compounded by the fact that over 60% of patients present with advanced disease stages III or IV (Jedy-Agba et al, 2016). In these resource constrained environments, chemotherapy often serves as the primary systematic intervention. However, pathological complete response rates remain highly variable, ranging from 10% to 40% (Cortazar et al, 2014; Spring et al, 2020). The lack of access to genomic profiling, currently available in fewer than 15% of regional oncology centers, leaves clinicians without the tools necessary to predict treatment sensitivity effectively (WHO 2020, Vanderpuye et al, 2017).

Background and Rationale

To address these systematic gaps, we have developed an interpretable artificial intelligence model specifically tailored to African healthcare realities. Unlike traditional "black box" algorithms our approach focuses on transparency by integrating routinely available clinical variables including tumor stage, histological subtype and receptor status with imaging features derived from standard diagnostic scans. The system generates a quantified probability of tumor shrinkage and treatment resistance, providing a clear breakdown of the clinical drivers behind each prediction.

Interpretability is not merely a technical preference but a clinical necessity in environments with high patient volumes and limited diagnostic infrastructure. By ensuring that AI outputs can be understood and validated by oncologists in real time, we align our technology with global ethical standards for healthcare (WHO 2021; Topol, 2019). This transparency fosters the trust required for clinical adoption supports rigorous evidence based decision making (Holzinger et al, 2017).

Our long term objective is to bridge the global oncology divide by creating precision medical tools adapted for African populations. By focussing on robustness in low data environments and diverse genetic backgrounds, we aim to improve treatment efficiency and patient outcomes. Identifying non-responders early allows for the optimization of limited chemotherapy resources and the timely consideration of alternative referral pathways, ultimately working toward a more equitable future in global cancer care.

Selected References

• Bray, F et al (2018). Global cancer statistics 2018. CA A Cancer Journal for Clinicians.
• Cortazar, P et al (2014) Pathological complete response in breast cancer. CTNeoBC pooled analysis. Lancet Oncology.
• Holzinger, A et al (2017) Explainable AI in medicine. Knowledge Based Systems.
• Jedy Agba, E et al (2016) Cancer incidence in sub-Saharan Africa. Lancet Global Health.
• Spring, L.M et al (2020) pCR and survival outcomes in breast cancer. Clinical Cancer Research.
• Sung, H et al (2021) Global Cancer Statistics 2020. CA: A Cancer Journal for Clinicians.
• Topol, E (2019) High performance medicine and AI in healthcare. Nature Medicine.
• Vanderpuye, V et al (2017) Cancer care in Africa, challenges and opportunities. The Lancet Oncology.
• WHO (2021). Ethics and governance of artificial intelligence for health, World Health Organization.

INVESTIGATING CHEMOTHERAPY RESISTANCE
IN BREAST CANCER

Dr. Shamim Nabuuma Kaliisa, MRes Data Science for Health

As a researcher, I have examined numerous medical publications to understand the systematic failure of breast cancer to respond to therapeutic interventions. My background in a Master of Research in patient datasets⁵. My findings indicate that chemotherapy resistance is far from a singular mechanical failure, rather it is a sophisticated, multi-layered defensive strategy orchestrated by the malignancy itself^5,6.

Current data suggests that between 30% and 50% of patients will experience treatment resistance during their clinical journey⁵. This phenomenon stems from the inherent heterogeneity of tumors, which contain diverse cell populations. While standard chemotherapy may successfully eliminate the majority of malignant cells, specialized resistant populations survive and initiate tumor regrowth — a process fueled by both pre-existing intrinsic traits and treatment induced acquired factors ^5,6. These cells employ molecular pumps, specifically from the ATP binding casette (ABC) transporter family including P-gp and BCRP, to actively expel pharmaceutical agents from the intracellular environment before they can exert a therapeutic effect^5,6. Furthermore, tumors produce intracellular enzymes such as glutathione S-transferase, that chemically neutralize drugs. Malignant cells may also alter specific molecular targets, such as β-tubulin or utilize sophiscated DNA repair pathways to counteract chemotherapy induced damage^5,6.

Beyond cellular mechanics, the tumor microenvironment serves as a critical sanctuary. Hypoxic conditions activate the HIF1α protein, which facilitates metabolic reprogramming and sustained drug resistance^5,6. Additionally, specialized cancer stem cells, identified by CD44+/CD24- markers, remain dormant during active treatment to regenerate the tumor at a later stage^5,6. Perhaps most strikingly, resistant cells utilize exosomes to transmit functional proteins and genetic material, effectively training surrounding cells to survive therapeutic pressure^5,6. My research is dedicated to mapping these survival pathways to improve our ability to predict resistance and develop precision strategies to dismantle these biological defenses⁵.

References

• Al-Hajji, M et al (2003) Prospective identification of tumorigenenic breast cancer cells. PNAS.
• Robey, R.W et al (2018) Revisiting the role of ABC transporters in multidrug resistant cancer. Nature Reviews Cancer.
• Zhang, Y et al (2024) Hypoxia-inducible factor in breast cancer: role and target. Frontiers in Immunology.
• Xiong, Y et al (2025) Emerging strategies to overcome resistance in breast cancer. Cancers.
• Research Analysis by Dr. Shamim Nabuuma Kaliisa (2026) Chil AI Lab & Chil Femtech Center.
• Scientific Facts: Chemotherapy Resistance in Breast Cancer Summary. Chil Group Precision Oncology Division.